Skip to main content

Scenario tables

One table per scenario and operating point, assembled from the stored cells. Rows are ordered by their ratio to the anchor. #n means the row is separated from every row above it by more than the larger drift of the two (for two rows from different batches, by more than the larger of row drift + its batch's anchor drift). ≈ means this machine did not establish where it belongs.

renders A / B is distinct renders over deliveries for the two views. If a library woke a consumer without changing what it shows, the difference counts as redundant deliveries, which are counted, not hidden.

batch insert batchInsertrepresentative

One step: every row is inserted into an empty store in one batch. Both views go from empty to a full page and must each re-render exactly once.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-array733.96 µs0.50×1.5%151231/1 · 1/12026-10-22busy, gate-raised
#2xstate-store998.13 µs0.68×2.2%152971/1 · 1/12026-10-22busy, gate-raised
#3zustand1.051 ms0.71×4.1%152451/1 · 1/12026-10-22busy, gate-raised
≈baseline-indexed3.147 ms1.00×44.2%152461/1 · 1/12026-10-31stale: harness, unresolved, busy, gate-raised
≈oimdb-sorted-index1.672 ms1.14×3.7%152851/1 · 1/12026-10-22stale: adapter, stale: subject, busy, gate-raised
≈tanstack-store1.754 ms1.19×31.0%152441/1 · 1/12026-10-22busy, gate-raised
≈effector1.954 ms1.33×7.0%153141/1 · 1/12026-10-22busy, gate-raised
≈nanostores2.273 ms1.54×1.9%152161/1 · 1/12026-10-22busy, gate-raised
≈nanostores-per-row3.193 ms2.17×53.0%152161/1 · 1/12026-10-22busy, gate-raised
≈vue-reactivity3.439 ms2.29×5.6%153321/1 · 1/12026-10-21gate-raised
≈preact-signals3.525 ms2.35×5.2%152981/1 · 1/12026-10-21gate-raised
≈oimdb3.868 ms2.38×27.5%152851/1 · 1/12026-10-31stale: adapter, stale: subject, unresolved, busy, gate-raised
≈alien-signals3.950 ms2.63×4.2%153281/1 · 1/12026-10-21gate-raised
≈solid-js5.102 ms3.40×115.1%153221/1 · 1/12026-10-21gate-raised
≈lokijs5.939 ms4.04×2.7%152121/1 · 1/12026-10-22busy, gate-raised
#16jotai14.143 ms9.61×13.1%153051/1 · 1/12026-10-22busy, gate-raised
≈oimdb-3.12.116.898 ms10×39.7%152161/1 · 1/12026-10-31unresolved, busy, gate-raised
≈redux-toolkit16.102 ms11×3.1%152901/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db19.130 ms13×15.4%152491/1 · 1/12026-10-22busy, gate-raised
≈mobx25.129 ms17×2.4%152001/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db-eager39.760 ms27×7.8%152491/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db-indexed40.573 ms28×9.3%152491/1 · 1/12026-10-22busy, gate-raised
≈valtio43.924 ms30×25.6%152121/1 · 1/12026-10-22busy, gate-raised
≈legend-state51.336 ms35×1.5%151821/1 · 1/12026-10-22busy, gate-raised
≈tinybase67.730 ms46×2.4%152141/1 · 1/12026-10-22busy, gate-raised
≈signaldb71.417 ms49×7.7%152711/1 · 1/12026-10-22busy, gate-raised

insert into a populated store insertIntoPopulatedrepresentative

One step: 100 new rows are inserted in one batch into a full store, with every group sorted and every visible row subscribed. They sort first, so they appear at the top of every group and both views must change. "batch insert" fills an empty store; this adds to one already in use.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈lokijs113.67 µs0.34×31.5%452121/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
#2oimdb-sorted-index244.21 µs0.73×6.4%452851/1 · 1/12026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb307.50 µs0.87×14.6%452851/1 · 1/12026-10-31stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈tanstack-store292.04 µs0.87×5.1%452441/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈vue-reactivity294.92 µs0.88×4.8%453321/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈preact-signals304.29 µs0.91×14.9%452981/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈alien-signals312.92 µs0.93×4.2%453281/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈nanostores-per-row332.13 µs0.99×15.7%452161/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈baseline-indexed365.13 µs1.00×75.9%452461/1 · 1/12026-10-31stale: harness, unresolved, busy, gate-raised
≈baseline-array347.88 µs1.04×2.8%451231/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈solid-js418.46 µs1.25×8.8%453221/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈jotai585.25 µs1.74×10.5%453051/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈oimdb-3.12.1803.54 µs2.26×144.8%452161/1 · 1/12026-10-31fine-unresolved, busy, gate-raised
≈effector839.13 µs2.50×2.7%453141/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈mobx902.17 µs2.69×5.6%452001/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈valtio1.148 ms3.42×13.1%452121/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈tinybase1.629 ms4.85×12.7%452141/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈legend-state2.074 ms6.18×20.7%451821/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈zustand2.315 ms6.90×1.3%452451/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed2.318 ms6.91×7.5%452491/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈xstate-store2.335 ms6.96×1.6%452971/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db2.664 ms7.94×16.6%452491/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager2.849 ms8.49×45.6%452491/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
#24redux-toolkit5.103 ms15×2.1%452901/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈signaldb7.496 ms22×3.6%452711/1 · 1/12026-10-22fine-unresolved, busy, gate-raised
≈nanostores7.560 ms23×1.2%452161/1 · 1/12026-10-22fine-unresolved, busy, gate-raised

the population keeps growing growUnderSubscriptionnarrow

8 steps. Each inserts a batch of new rows spread over every group and sorted first, and nothing is deleted, so the store ends forty per cent larger than it started. A store grown step by step is not the same as one loaded in one go: sorted lists and indexes can cost more as they grow. Every batch shows at the top of both views.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-indexed835.58 µs1.00×1.4%272468/8 · 8/82026-10-22busy, gate-raised
#2oimdb-sorted-index1.082 ms1.30×3.1%272858/8 · 8/82026-10-22stale: adapter, stale: subject, busy, gate-raised
≈tanstack-store1.221 ms1.46×31.7%272448/8 · 8/82026-10-22busy, gate-raised
≈oimdb1.255 ms1.50×1.0%272858/8 · 8/82026-10-22stale: adapter, stale: subject, busy, gate-raised
≈baseline-array1.384 ms1.66×0.1%271238/8 · 8/82026-10-22busy, gate-raised
≈lokijs1.512 ms1.81×6.1%272128/8 · 8/82026-10-22busy, gate-raised
#7preact-signals1.757 ms2.10×5.2%272988/8 · 8/82026-10-22busy, gate-raised
≈vue-reactivity1.962 ms2.35×12.2%273328/8 · 8/82026-10-22busy, gate-raised
≈alien-signals2.502 ms2.99×9.6%273288/8 · 8/82026-10-22busy, gate-raised
#10nanostores-per-row3.002 ms3.59×4.7%272168/8 · 8/82026-10-22busy, gate-raised
≈solid-js3.700 ms4.43×3.7%273228/8 · 8/82026-10-22busy, gate-raised
#12effector6.223 ms7.45×6.1%273148/8 · 8/82026-10-22busy, gate-raised
#13oimdb-3.12.18.350 ms9.99×2.0%272168/8 · 8/82026-10-22busy, gate-raised
#14jotai9.628 ms12×2.0%273058/8 · 8/82026-10-22busy, gate-raised
#15mobx12.508 ms15×6.5%272008/8 · 8/82026-10-22busy, gate-raised
#16tanstack-db17.120 ms20×1.1%272498/8 · 8/82026-10-22busy, gate-raised
#17xstate-store21.695 ms26×0.4%272978/8 · 8/82026-10-22busy, gate-raised
≈zustand21.858 ms26×0.8%272458/8 · 8/82026-10-22busy, gate-raised
#19nanostores23.447 ms28×2.7%272168/8 · 8/82026-10-22busy, gate-raised
≈tanstack-db-indexed26.068 ms31×18.8%272498/8 · 8/82026-10-22busy, gate-raised
≈tanstack-db-eager26.769 ms32×0.6%272498/8 · 8/82026-10-22busy, gate-raised
≈legend-state30.390 ms36×2.5%271828/8 · 8/82026-10-22busy, gate-raised
≈tinybase30.710 ms37×1.3%272148/8 · 8/82026-10-22busy, gate-raised
≈valtio36.122 ms43×20.4%272128/8 · 8/82026-10-22busy, gate-raised
#25redux-toolkit52.251 ms63×3.5%272908/8 · 8/82026-10-22busy, gate-raised
≈signaldb109.912 ms132×0.7%272718/8 · 8/82026-10-22busy, gate-raised

cold start to first render coldStartnarrow

One step, timed from nothing: create the store and both views, subscribe them, insert every row, and render the first page of each view. The only scenario that times creating the store. Compare with "batch insert", the same work without creating the store: the difference is what setting up a store and its subscriptions costs.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array733.42 µs0.49×1.9%151231/1 · 1/12026-10-22gate-raised
≈xstate-store1.036 ms0.69×0.8%152971/1 · 1/12026-10-22gate-raised
#3zustand1.101 ms0.74×3.4%152451/1 · 1/12026-10-22gate-raised
≈baseline-indexed1.814 ms1.00×210.3%152461/1 · 1/12026-10-31stale: harness, unresolved, busy, gate-raised
≈oimdb-sorted-index1.671 ms1.12×2.2%152851/1 · 1/12026-10-22stale: adapter, stale: subject, gate-raised
≈tanstack-store1.744 ms1.17×24.0%152441/1 · 1/12026-10-22gate-raised
≈effector2.022 ms1.35×1.9%153142/2 · 2/22026-10-22gate-raised
≈oimdb2.218 ms1.44×14.9%152851/1 · 1/12026-10-31stale: adapter, stale: subject, busy, gate-raised
≈nanostores2.195 ms1.47×1.6%152162/2 · 2/22026-10-22gate-raised
≈nanostores-per-row3.077 ms2.06×33.2%152162/2 · 2/22026-10-22gate-raised
≈preact-signals3.706 ms2.50×36.2%152982/2 · 2/22026-10-21busy, gate-raised
≈alien-signals3.798 ms2.56×11.9%153282/2 · 2/22026-10-21busy, gate-raised
≈vue-reactivity4.260 ms2.87×17.2%153322/2 · 2/22026-10-21busy, gate-raised
≈solid-js4.866 ms3.28×27.2%153222/2 · 2/22026-10-21busy, gate-raised
≈lokijs5.877 ms3.93×2.2%152121/1 · 1/12026-10-22gate-raised
≈oimdb-3.12.19.010 ms5.86×21.6%152161/1 · 1/12026-10-31busy, gate-raised
#17jotai14.451 ms9.66×6.6%153051/1 · 1/12026-10-22gate-raised
#18redux-toolkit16.084 ms11×2.6%152901/1 · 1/12026-10-22gate-raised
#19mobx20.661 ms14×10.8%152002/2 · 2/22026-10-22busy, gate-raised
≈tanstack-db24.604 ms16×30.8%152491/1 · 1/12026-10-22gate-raised
≈tanstack-db-eager39.380 ms26×5.9%152491/1 · 1/12026-10-22gate-raised
≈tanstack-db-indexed39.470 ms26×34.1%152491/1 · 1/12026-10-22gate-raised
≈valtio45.997 ms31×2.1%152121/1 · 1/12026-10-22gate-raised
≈legend-state51.310 ms34×2.5%151822/2 · 2/22026-10-22gate-raised
#25tinybase70.752 ms47×2.6%152141/1 · 1/12026-10-22gate-raised
≈signaldb70.830 ms47×0.7%152711/1 · 1/12026-10-22gate-raised

tear the screen down tearDownnarrow

One step: the whole screen, a full store with both views and every subscription, is disposed. The teardown itself must render nothing. Afterwards, outside the timing, rows are written into the disposed screen and nothing may render, so a dispose that disposed nothing fails instead of looking fastest. A library that refuses that write also passes.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈xstate-store2.75 µs0.51×10.6%452970/0 · 0/02026-10-22unresolved, busy, gate-raised
≈zustand3.08 µs0.57×41.9%452450/0 · 0/02026-10-22unresolved, busy, gate-raised
≈baseline-array3.83 µs0.71×21.8%451230/0 · 0/02026-10-22unresolved, busy, gate-raised
≈oimdb-sorted-index5.25 µs0.98×37.3%452850/0 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈baseline-indexed5.38 µs1.00×29.5%452460/0 · 0/02026-10-22unresolved, busy, gate-raised
≈oimdb6.71 µs1.25×10.5%452850/0 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈redux-toolkit7.13 µs1.33×26.3%452900/0 · 0/02026-10-22unresolved, busy, gate-raised
≈alien-signals8.13 µs1.38×24.1%453280/0 · 0/02026-10-22unresolved, busy, gate-raised
≈preact-signals10.00 µs1.70×20.4%452980/0 · 0/02026-10-22unresolved, busy, gate-raised
≈vue-reactivity10.21 µs1.74×5.3%453320/0 · 0/02026-10-22unresolved, busy, gate-raised
≈tanstack-store11.00 µs2.05×8.7%452440/0 · 0/02026-10-22unresolved, busy, gate-raised
≈effector11.79 µs2.19×32.1%453140/0 · 0/02026-10-22unresolved, busy, gate-raised
≈solid-js14.63 µs2.49×22.8%453220/0 · 0/02026-10-22unresolved, busy, gate-raised
≈nanostores13.71 µs2.55×48.6%452160/0 · 0/02026-10-22unresolved, busy, gate-raised
≈nanostores-per-row21.67 µs4.03×3.8%452160/0 · 0/02026-10-22unresolved, busy, gate-raised
≈lokijs43.71 µs8.13×7.7%452120/0 · 0/02026-10-22unresolved, busy, gate-raised
≈oimdb-3.12.149.71 µs9.25×4.4%452160/0 · 0/02026-10-22unresolved, busy, gate-raised
≈legend-state59.04 µs11×3.0%451820/0 · 0/02026-10-22unresolved, busy, gate-raised
≈tanstack-db70.67 µs13×35.1%452490/0 · 0/02026-10-22unresolved, busy, gate-raised
≈tanstack-db-indexed96.83 µs18×25.9%452490/0 · 0/02026-10-22unresolved, busy, gate-raised
≈tanstack-db-eager107.38 µs20×8.5%452490/0 · 0/02026-10-22unresolved, busy, gate-raised
≈valtio177.88 µs33×19.3%452120/0 · 0/02026-10-22unresolved, busy, gate-raised
≈mobx198.25 µs34×15.0%452000/0 · 0/02026-10-22unresolved, busy, gate-raised
≈signaldb278.58 µs52×2.5%452710/0 · 0/02026-10-22unresolved, busy, gate-raised
#25jotai301.88 µs56×7.1%453050/0 · 0/02026-10-22unresolved, busy, gate-raised
#26tinybase15.173 ms2823×1.5%452140/0 · 0/02026-10-22unresolved, busy, gate-raised

teardown lower bound tearDownFloorsynthetic

Lower bound, no library: a hand-written screen (every row in a Map, two views and one subscriber per visible row) is discarded by dropping the single reference to it. No library can tear a screen down for less. Afterwards, outside the timing, rows are written into the discarded screen and nothing may render.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-array1.46 µs17.1%451230/0 · 0/02026-10-22busy, gate-raised

patch one column patchColumnrepresentative

32 steps. Each step changes the `label` of one row visible in view A. View A must re-render with the new text every step. View B shows a different group and must not change; how often it was woken anyway is counted.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈oimdb-3.12.1146.54 µs0.26×76.7%43521632/32 · 0/02026-10-31unresolved, busy, gate-raised
≈oimdb-sorted-index133.52 µs0.37×1.0%43528532/32 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb152.75 µs0.42×4.2%43528532/32 · 0/02026-10-31fine-unresolved, busy, gate-raised
≈alien-signals154.22 µs0.43×0.5%43532832/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#5mobx229.10 µs0.64×3.2%43520032/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈preact-signals233.75 µs0.65×0.8%43529832/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#7vue-reactivity252.37 µs0.70×1.4%43533232/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#8nanostores-per-row300.00 µs0.84×1.0%43521632/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#9solid-js351.66 µs0.98×1.5%43532232/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈lokijs356.10 µs0.99×0.8%43521232/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈baseline-indexed366.16 µs1.00×1.3%43524632/32 · 0/02026-10-31fine-unresolved, busy, gate-raised
#12nanostores377.22 µs1.05×1.9%43521632/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#13tanstack-store435.49 µs1.21×1.9%43524432/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tinybase557.01 µs1.55×2.5%43521432/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈effector573.60 µs1.60×3.6%43531432/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db1.461 ms4.08×6.2%43524932/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager1.483 ms4.14×10.8%43524932/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed1.484 ms4.14×2.8%43524932/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#19valtio2.757 ms7.69×9.3%43521232/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20baseline-array3.714 ms10×0.2%43512332/32 · 0/322026-10-22fine-unresolved, busy, gate-raised
#21jotai5.384 ms15×2.7%43530532/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#22legend-state14.792 ms41×2.8%43518232/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23xstate-store69.899 ms195×0.5%43529732/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈zustand69.963 ms195×0.8%43524532/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit104.737 ms292×0.4%43529032/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb186.532 ms520×4.8%43527132/32 · 0/02026-10-22fine-unresolved, busy, gate-raised

patch a row off screen patchOutsideWindowrepresentative

32 steps. Each step changes the `label` of a row in view A's group that sits just below the visible page: in the group, but not on screen. Neither view may re-render. Measured: how often a library wakes a view for a change nobody can see.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1preact-signals1.14 µs0.65×5.1%5852980/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
#2alien-signals1.24 µs0.70×5.0%5853280/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈vue-reactivity1.28 µs0.73×8.6%5853320/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈solid-js1.38 µs0.79×3.2%5853220/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈baseline-indexed2.19 µs1.00×171.4%5852460/0 · 0/02026-10-31stale: harness, fine-unresolved, busy, gate-raised
≈oimdb2.69 µs1.23×29.0%5852850/0 · 0/02026-10-31stale: harness, stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb-sorted-index3.02 µs1.65×1.6%5852850/0 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb-3.12.16.33 µs3.46×3.4%5852160/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈mobx6.40 µs3.50×11.7%5852000/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#10jotai234.06 µs128×7.4%5853050/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#11nanostores-per-row293.25 µs160×0.4%5852160/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tinybase294.38 µs161×5.2%5852140/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#13lokijs352.73 µs193×3.1%5852120/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈nanostores364.89 µs200×11.5%5852160/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#15tanstack-store425.25 µs233×3.3%5852440/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#16effector562.85 µs308×4.2%5853140/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#17tanstack-db722.94 µs395×7.6%5852490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager758.05 µs415×5.1%5852490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed775.42 µs424×14.4%5852490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20valtio2.726 ms1491×0.5%5852120/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#21baseline-array3.676 ms2011×0.6%5851230/32 · 0/322026-10-22fine-unresolved, busy, gate-raised
#22legend-state15.121 ms8272×1.4%5851820/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23xstate-store70.597 ms38620×3.5%5852970/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈zustand70.746 ms38701×2.1%5852450/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit104.541 ms57189×1.1%5852900/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb171.602 ms93874×5.3%5852710/0 · 0/02026-10-22fine-unresolved, busy, gate-raised

patch an unread column patchNotenarrow

32 steps. The same rows and the same calls as "patch one column", but the changed field is `note`, which no view displays. Neither view may re-render. The difference between the two scenarios is what tracking changes per field saves.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed1.01 µs1.00×7.3%6302460/0 · 0/02026-10-31stale: harness, fine-unresolved, busy, gate-raised
#2preact-signals1.20 µs1.46×8.9%6302980/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈alien-signals1.27 µs1.56×8.1%6303280/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈vue-reactivity1.29 µs1.58×9.3%6303320/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈solid-js1.42 µs1.73×9.2%6303220/0 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈oimdb-sorted-index2.33 µs2.82×14.8%6302850/0 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb3.93 µs3.87×10.9%6302850/0 · 0/02026-10-31stale: harness, stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#8mobx6.41 µs7.78×5.0%6302000/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#9legend-state87.47 µs106×2.7%6301820/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#10oimdb-3.12.1134.41 µs163×0.5%6302160/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#11tinybase230.95 µs280×7.0%6302140/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#12jotai276.63 µs336×2.9%6303050/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#13nanostores-per-row303.59 µs368×1.9%6302160/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#14lokijs355.87 µs432×1.0%6302120/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#15nanostores372.53 µs452×4.6%6302160/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#16tanstack-store430.15 µs522×7.2%6302440/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#17effector574.82 µs697×2.9%6303140/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db1.101 ms1336×12.1%6302490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager1.106 ms1342×6.2%6302490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed1.148 ms1392×16.7%6302490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#21valtio2.759 ms3346×5.1%6302120/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#22baseline-array3.785 ms4592×2.2%6301230/32 · 0/322026-10-22fine-unresolved, busy, gate-raised
#23xstate-store70.478 ms85490×1.0%6302970/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈zustand71.854 ms87159×2.2%6302450/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit106.065 ms128656×2.3%6302900/32 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb191.274 ms232015×5.4%6302710/32 · 0/02026-10-22fine-unresolved, busy, gate-raised

batch patch batchPatchrepresentative

One step: the `label` of every tenth row in the whole store (at least four pages of rows, spread over every group) is changed inside one batch, and the changes are delivered once. This is a sync response or a bulk edit. Most changes land in groups no view shows. Both views must end showing the new text. How many times each re-renders is up to the library: one with a batch API renders once per view, one without renders once per changed visible row, so only the final screen is checked.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈vue-reactivity69.54 µs0.98×3.1%3003321/1 · 1/12026-10-21busy, gate-raised
≈alien-signals70.46 µs0.99×3.6%3003281/1 · 1/12026-10-21busy, gate-raised
≈baseline-indexed71.37 µs1.00×7.0%3002461/1 · 1/12026-10-22busy, gate-raised
#4oimdb-sorted-index94.83 µs1.33×2.1%3002851/1 · 1/12026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb95.35 µs1.34×6.4%3002851/1 · 1/12026-10-22stale: adapter, stale: subject, busy, gate-raised
≈preact-signals100.48 µs1.41×6.1%3002981/1 · 1/12026-10-21busy, gate-raised
#7solid-js117.67 µs1.66×2.9%3003221/1 · 1/12026-10-21busy, gate-raised
#8tanstack-store151.76 µs2.13×6.5%3002441/1 · 1/12026-10-22busy, gate-raised
≈baseline-array152.34 µs2.13×0.2%3001231/1 · 1/12026-10-22busy, gate-raised
#10oimdb-3.12.1200.63 µs2.81×1.9%3002161/1 · 1/12026-10-22busy, gate-raised
#11nanostores-per-row324.88 µs4.55×4.6%3002161/1 · 1/12026-10-22busy, gate-raised
#12mobx344.20 µs4.82×5.4%3002001/1 · 1/12026-10-22busy, gate-raised
#13lokijs449.32 µs6.30×3.3%3002121/1 · 1/12026-10-22busy, gate-raised
#14valtio511.71 µs7.17×2.1%3002121/1 · 1/12026-10-22busy, gate-raised
#15effector669.14 µs9.38×5.9%3003141/1 · 1/12026-10-22busy, gate-raised
#16jotai1.286 ms18×4.2%3003051/1 · 1/12026-10-22busy, gate-raised
#17tinybase2.071 ms29×10.9%3002141/1 · 1/12026-10-22busy, gate-raised
#18zustand2.299 ms32×1.9%3002451/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db2.922 ms41×8.6%3002491/1 · 1/12026-10-22busy, gate-raised
≈legend-state2.970 ms42×9.5%3001821/1 · 1/12026-10-22busy, gate-raised
#21tanstack-db-indexed3.363 ms47×8.4%3002491/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db-eager3.512 ms49×4.3%3002491/1 · 1/12026-10-22busy, gate-raised
#23redux-toolkit4.314 ms60×4.4%3002901/1 · 1/12026-10-22busy, gate-raised
#24nanostores15.630 ms219×3.2%3002161/1 · 1/12026-10-22busy, gate-raised
#25signaldb97.979 ms1373×4.0%3002711/1 · 1/12026-10-22busy, gate-raised
#26xstate-store2228.614 ms31225×1.2%3002971/1 · 1/12026-10-22busy, gate-raised

batch patch an unread column batchPatchNotesynthetic

One step: the same rows as "batch patch" are changed in one batch, but in `note`, a field no view displays. Neither view may show new text; any re-render is wasted work and is counted. Against "batch patch" it shows the cost of a batch nobody sees; against "patch an unread column" it shows what batching saves. The changes are checked in the store afterwards.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed37.69 µs1.00×0.4%3002460/0 · 0/02026-10-22busy, gate-raised
#2alien-signals56.26 µs1.50×3.6%3003280/0 · 0/02026-10-21busy, gate-raised
≈vue-reactivity56.82 µs1.52×2.4%3003320/0 · 0/02026-10-21busy, gate-raised
#4oimdb-sorted-index64.34 µs1.72×6.9%3002850/0 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈solid-js65.94 µs1.76×2.5%3003220/0 · 0/02026-10-21busy, gate-raised
≈oimdb66.51 µs1.77×10.7%3002850/0 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈preact-signals69.50 µs1.86×14.6%3002980/0 · 0/02026-10-21busy, gate-raised
#8tanstack-store151.08 µs4.01×3.1%3002440/1 · 0/12026-10-22busy, gate-raised
≈baseline-array150.91 µs4.02×3.1%3001230/1 · 0/12026-10-22busy, gate-raised
#10oimdb-3.12.1195.17 µs5.20×3.2%3002160/1 · 0/12026-10-22busy, gate-raised
#11mobx258.63 µs6.89×13.3%3002000/0 · 0/02026-10-22busy, gate-raised
#12nanostores-per-row332.18 µs8.85×17.6%3002160/1 · 0/12026-10-22busy, gate-raised
#13lokijs445.98 µs12×1.6%3002120/1 · 0/12026-10-22busy, gate-raised
#14valtio526.14 µs14×11.2%3002120/1 · 0/12026-10-22busy, gate-raised
#15effector675.06 µs18×8.3%3003140/1 · 0/12026-10-22busy, gate-raised
#16jotai848.72 µs23×7.1%3003050/0 · 0/02026-10-22busy, gate-raised
#17legend-state1.802 ms48×17.4%3001820/0 · 0/02026-10-22busy, gate-raised
≈tinybase1.952 ms52×11.9%3002140/0 · 0/02026-10-22busy, gate-raised
#19zustand2.359 ms63×3.0%3002450/1 · 0/12026-10-22busy, gate-raised
≈tanstack-db2.968 ms79×13.4%3002490/0 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager3.352 ms89×7.8%3002490/0 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed3.384 ms90×8.9%3002490/0 · 0/02026-10-22busy, gate-raised
#23redux-toolkit4.320 ms115×1.7%3002900/1 · 0/12026-10-22busy, gate-raised
#24nanostores15.669 ms418×3.9%3002160/1 · 0/12026-10-22busy, gate-raised
#25signaldb96.770 ms2580×2.5%3002710/1 · 0/12026-10-22busy, gate-raised
#26xstate-store2263.407 ms60057×0.8%3002970/1 · 0/12026-10-22busy, gate-raised

most rows change every frame patchMostRowsnarrow

2 steps. Each step changes the `label` of one row in 2 across the whole store, the same share of every group, and delivers once. Every other write scenario changes a handful of rows; this one checks the opposite end, where most of what is subscribed changes at once. Both views must end every step showing new text.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed632.88 µs1.00×58.6%152462/2 · 2/22026-10-22fine-unresolved, busy, gate-raised
≈baseline-array632.13 µs1.05×43.5%151232/2 · 2/22026-10-22unresolved, busy, gate-raised
≈alien-signals901.46 µs1.08×15.9%153282/2 · 2/22026-10-21unresolved, busy, gate-raised
≈vue-reactivity1.045 ms1.25×7.5%153322/2 · 2/22026-10-21unresolved, busy, gate-raised
≈preact-signals1.155 ms1.38×3.9%152982/2 · 2/22026-10-21unresolved, busy, gate-raised
≈solid-js1.264 ms1.51×1.3%153222/2 · 2/22026-10-21unresolved, busy, gate-raised
≈oimdb1.304 ms2.17×3.6%152852/2 · 2/22026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈oimdb-sorted-index1.319 ms2.19×6.1%152852/2 · 2/22026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈tanstack-store1.397 ms2.21×17.4%152442/2 · 2/22026-10-22fine-unresolved, busy, gate-raised
#10oimdb-3.12.12.405 ms4.00×10.9%152162/2 · 2/22026-10-22unresolved, busy, gate-raised
#11effector3.210 ms5.34×1.6%153142/2 · 2/22026-10-22unresolved, busy, gate-raised
≈nanostores-per-row3.442 ms5.73×9.6%152162/2 · 2/22026-10-22unresolved, busy, gate-raised
≈mobx3.571 ms5.94×43.5%152002/2 · 2/22026-10-22unresolved, busy, gate-raised
≈lokijs4.346 ms7.23×8.5%152122/2 · 2/22026-10-22unresolved, busy, gate-raised
≈valtio5.502 ms8.69×11.8%152122/2 · 2/22026-10-22fine-unresolved, busy, gate-raised
≈zustand5.706 ms9.02×0.8%152452/2 · 2/22026-10-22fine-unresolved, busy, gate-raised
#17jotai10.629 ms18×11.5%153052/2 · 2/22026-10-22unresolved, busy, gate-raised
≈redux-toolkit18.491 ms29×1.6%152902/2 · 2/22026-10-22fine-unresolved, busy, gate-raised
≈legend-state23.270 ms39×5.0%151822/2 · 2/22026-10-22unresolved, busy, gate-raised
≈tinybase23.982 ms40×1.6%152142/2 · 2/22026-10-22unresolved, busy, gate-raised
#21tanstack-db45.518 ms76×15.6%152492/2 · 2/22026-10-22unresolved, busy, gate-raised
≈tanstack-db-indexed49.499 ms82×9.3%152492/2 · 2/22026-10-22unresolved, busy, gate-raised
≈tanstack-db-eager50.843 ms85×5.1%152492/2 · 2/22026-10-22unresolved, busy, gate-raised
≈nanostores166.189 ms276×0.6%152162/2 · 2/22026-10-22unresolved, busy, gate-raised
#25signaldb1105.348 ms1839×4.4%152712/2 · 2/22026-10-22unresolved, busy, gate-raised
#26xstate-store22553.985 ms35637×0.4%152972/2 · 2/22026-10-22fine-unresolved, busy, gate-raised

one frame, every kind of change mixedFramerepresentative

One step with five kinds of change delivered together: 4 new rows inserted at the top of view A, its last 4 rows deleted, `label` changed on rows in between, 4 rows moved to group B, and a few rows re-sorted inside view A. Each kind touches different rows. This is what one message from a server can carry; every other scenario has one kind. Both views must end full and correct, with a new row at the top of view A.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed94.42 µs1.00×7.9%272461/1 · 1/12026-10-22busy, gate-raised
≈vue-reactivity94.79 µs1.00×6.3%273321/1 · 1/12026-10-22busy, gate-raised
≈alien-signals94.96 µs1.01×7.4%273281/1 · 1/12026-10-22busy, gate-raised
≈preact-signals100.71 µs1.07×5.1%272981/1 · 1/12026-10-22busy, gate-raised
≈effector109.92 µs1.16×10.7%273141/1 · 1/12026-10-22busy, gate-raised
≈tanstack-store114.54 µs1.21×15.1%272441/1 · 1/12026-10-22busy, gate-raised
≈oimdb-sorted-index116.33 µs1.28×5.3%272851/1 · 1/12026-10-22stale: adapter, stale: subject, busy, gate-raised
≈solid-js121.50 µs1.29×12.6%273221/1 · 1/12026-10-22busy, gate-raised
#9oimdb143.25 µs1.58×9.8%272851/1 · 1/12026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores-per-row151.79 µs1.61×27.0%272161/2 · 1/12026-10-22busy, gate-raised
≈baseline-array191.04 µs2.10×2.9%271231/1 · 1/12026-10-22busy, gate-raised
≈valtio399.75 µs4.23×4.4%272121/1 · 1/12026-10-22busy, gate-raised
#13oimdb-3.12.1468.83 µs5.16×10.0%272161/1 · 1/12026-10-22busy, gate-raised
≈mobx496.88 µs5.47×44.4%272001/1 · 1/12026-10-22busy, gate-raised
≈jotai679.17 µs7.47×29.4%273051/1 · 1/12026-10-22busy, gate-raised
≈nanostores720.54 µs7.63×1.5%272161/1 · 1/12026-10-22busy, gate-raised
≈tinybase918.63 µs10×5.0%272141/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db1.048 ms12×25.6%272491/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db-eager1.223 ms13×4.2%272491/1 · 1/12026-10-22busy, gate-raised
≈tanstack-db-indexed1.231 ms14×5.8%272491/1 · 1/12026-10-22busy, gate-raised
≈legend-state1.714 ms19×0.5%271821/1 · 1/12026-10-22busy, gate-raised
#22lokijs2.169 ms24×1.6%272121/1 · 1/12026-10-22busy, gate-raised
≈zustand2.330 ms25×1.1%272451/1 · 1/12026-10-22busy, gate-raised
#24redux-toolkit5.374 ms57×1.7%272901/1 · 1/12026-10-22busy, gate-raised
#25signaldb10.386 ms114×1.1%272711/1 · 1/12026-10-22busy, gate-raised
#26xstate-store28.842 ms305×0.3%272971/1 · 1/12026-10-22busy, gate-raised

row moves between groups moveGroupnarrow

8 steps. Each step moves one row from group A to group B or back. The moved rows sort first, so every move shows at the top of both views: view A loses a row and pulls in the next one from below, view B gains it.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index101.76 µs0.36×2.8%6302858/8 · 8/82026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#2preact-signals119.52 µs0.43×6.2%6302988/8 · 8/82026-10-21fine-unresolved, busy, gate-raised
#3alien-signals130.60 µs0.47×0.2%6303288/8 · 8/82026-10-21fine-unresolved, busy, gate-raised
#4vue-reactivity135.47 µs0.48×1.4%6303328/8 · 8/82026-10-21fine-unresolved, busy, gate-raised
≈solid-js157.54 µs0.56×2.1%6303228/8 · 8/82026-10-21fine-unresolved, busy, gate-raised
#6oimdb175.09 µs0.63×4.7%6302858/8 · 8/82026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈lokijs180.13 µs0.64×2.3%6302128/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈nanostores-per-row188.33 µs0.67×1.0%6302168/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#9tanstack-store209.05 µs0.74×4.5%6302448/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#10baseline-indexed282.45 µs1.00×2.1%6302468/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈oimdb-3.12.1285.36 µs1.01×6.1%6302168/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#12effector318.83 µs1.13×2.3%6303148/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#13nanostores587.25 µs2.08×2.3%6302168/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#14tanstack-db-eager638.64 µs2.26×8.1%6302498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed646.58 µs2.29×15.3%6302498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db649.06 µs2.30×3.8%6302498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#17tinybase932.53 µs3.30×1.7%6302148/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈baseline-array932.64 µs3.30×0.3%6301238/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#19valtio1.342 ms4.75×3.1%6302128/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#20mobx1.808 ms6.40×1.1%6302008/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#21jotai2.376 ms8.41×3.7%6303058/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#22legend-state7.782 ms28×2.2%6301828/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#23zustand18.062 ms64×0.6%6302458/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈xstate-store18.090 ms64×0.3%6302978/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit27.464 ms97×0.2%6302908/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#26signaldb53.080 ms188×1.9%6302718/8 · 8/82026-10-22fine-unresolved, busy, gate-raised

a row arrives mid-order moveAndReordernarrow

4 steps. Each step changes one row's `group` and `rank` together, so it leaves the top of view A and lands in the middle of view B, not at its top. Compare with "row moves between groups", where the row always lands at position 0, the cheapest place to insert into a sorted list. Both views must change every step and stay full, and the final positions are checked.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈preact-signals80.79 µs0.53×7.8%272984/4 · 4/42026-10-21fine-unresolved, busy, gate-raised
≈oimdb-sorted-index89.83 µs0.58×1.5%272854/4 · 4/42026-10-22stale: adapter, stale: subject, fine-unresolved, gate-raised
≈alien-signals91.21 µs0.59×7.1%273284/4 · 4/42026-10-21fine-unresolved, busy, gate-raised
≈vue-reactivity91.88 µs0.60×12.5%273324/4 · 4/42026-10-21fine-unresolved, busy, gate-raised
≈solid-js117.50 µs0.77×9.5%273224/4 · 4/42026-10-21fine-unresolved, busy, gate-raised
≈oimdb130.08 µs0.83×3.1%272854/4 · 4/42026-10-22stale: adapter, stale: subject, fine-unresolved, gate-raised
≈tanstack-store130.00 µs0.83×8.2%272444/4 · 4/42026-10-22fine-unresolved, gate-raised
≈nanostores-per-row154.92 µs0.99×52.9%272164/4 · 4/42026-10-22fine-unresolved, gate-raised
≈baseline-indexed156.00 µs1.00×2.6%272464/4 · 4/42026-10-22fine-unresolved, gate-raised
≈lokijs166.63 µs1.07×5.7%272124/4 · 4/42026-10-22fine-unresolved, gate-raised
≈effector184.67 µs1.18×4.1%273144/4 · 4/42026-10-22fine-unresolved, gate-raised
#12oimdb-3.12.1278.63 µs1.79×4.7%272164/4 · 4/42026-10-22fine-unresolved, gate-raised
≈baseline-array485.33 µs3.11×1.5%271234/4 · 4/42026-10-22fine-unresolved, gate-raised
#14nanostores636.04 µs4.08×2.6%272164/4 · 4/42026-10-22fine-unresolved, gate-raised
≈valtio809.29 µs5.19×3.2%272124/4 · 4/42026-10-22fine-unresolved, gate-raised
≈tanstack-db857.96 µs5.50×1.8%272494/4 · 4/42026-10-22fine-unresolved, gate-raised
≈tanstack-db-eager910.21 µs5.83×14.9%272494/4 · 4/42026-10-22fine-unresolved, gate-raised
≈tanstack-db-indexed981.79 µs6.29×5.9%272494/4 · 4/42026-10-22fine-unresolved, gate-raised
≈mobx1.079 ms6.92×9.2%272004/4 · 4/42026-10-22fine-unresolved, gate-raised
≈tinybase1.415 ms9.07×37.9%272144/4 · 4/42026-10-22fine-unresolved, gate-raised
≈jotai1.594 ms10×0.9%273054/4 · 4/42026-10-22fine-unresolved, gate-raised
≈legend-state4.888 ms31×16.8%271824/4 · 4/42026-10-22fine-unresolved, gate-raised
#23zustand8.999 ms58×1.2%272454/4 · 4/42026-10-22fine-unresolved, gate-raised
#24redux-toolkit14.428 ms92×0.6%272904/4 · 4/42026-10-22fine-unresolved, gate-raised
≈xstate-store17.763 ms114×2.3%272974/4 · 4/42026-10-22fine-unresolved, gate-raised
#26signaldb26.516 ms170×2.5%272714/4 · 4/42026-10-22fine-unresolved, gate-raised

sort key moves sortKeyMoverepresentative

8 steps. Each step changes one row's `rank` so that it moves to another place inside view A. The same rows stay on screen; only their order changes. View B must not change.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1preact-signals43.82 µs0.50×1.1%8402988/8 · 0/02026-10-21fine-unresolved, busy, gate-raised
#2alien-signals47.94 µs0.55×1.0%8403288/8 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈oimdb-sorted-index48.98 µs0.55×1.7%8402858/8 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#4vue-reactivity66.54 µs0.76×1.3%8403328/8 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈nanostores-per-row69.51 µs0.79×4.6%8402168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#6solid-js78.00 µs0.89×6.3%8403228/8 · 0/02026-10-21fine-unresolved, busy, gate-raised
≈oimdb80.91 µs0.92×3.3%8402858/8 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, gate-raised
#8baseline-indexed88.29 µs1.00×0.7%8402468/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#9nanostores93.40 µs1.06×0.8%8402168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#10tanstack-store106.24 µs1.20×4.3%8402448/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#11lokijs120.55 µs1.37×6.5%8402128/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#12effector143.82 µs1.63×4.9%8403148/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#13oimdb-3.12.1192.66 µs2.18×0.5%8402168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#14tinybase445.95 µs5.05×1.9%8402148/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#15tanstack-db551.60 µs6.25×3.0%8402498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#16mobx676.45 µs7.66×6.0%8402008/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈valtio697.80 µs7.90×0.7%8402128/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#18tanstack-db-indexed825.35 µs9.35×3.9%8402498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager831.50 µs9.42×3.1%8402498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20baseline-array924.88 µs10×1.8%8401238/8 · 0/82026-10-22fine-unresolved, busy, gate-raised
#21jotai1.128 ms13×3.0%8403058/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#22legend-state3.767 ms43×2.2%8401828/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23xstate-store17.856 ms202×1.0%8402978/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈zustand17.976 ms204×0.9%8402458/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit26.956 ms305×0.8%8402908/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb50.235 ms569×2.1%8402718/8 · 0/02026-10-22fine-unresolved, busy, gate-raised

the whole window re-sorted resortWindownarrow

8 steps. Each step changes the `rank` of all 32 rows of view A at once, so the whole page is reordered while the same rows stay on it. Compare with "sort key moves", which moves one row: together they show how the cost of keeping a list sorted grows with how much of the order changed. View B must not change.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed93.44 µs1.00×1.9%2102468/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#2tanstack-store132.34 µs1.42×3.6%2102448/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#3preact-signals140.58 µs1.53×0.8%2102988/8 · 0/02026-10-21fine-unresolved, gate-raised
#4nanostores-per-row160.75 µs1.72×2.7%2102168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈oimdb171.73 µs1.87×8.7%2102858/8 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈effector178.44 µs1.91×5.6%2103148/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#7vue-reactivity202.61 µs2.21×0.8%2103328/8 · 0/02026-10-21fine-unresolved, gate-raised
#8oimdb-sorted-index224.61 µs2.40×3.3%2102858/8 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈alien-signals220.86 µs2.41×2.1%2103288/8 · 0/02026-10-21fine-unresolved, gate-raised
≈nanostores227.48 µs2.43×5.0%2102168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈lokijs240.08 µs2.57×1.2%2102128/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#12solid-js297.07 µs3.24×3.2%2103228/8 · 0/02026-10-21fine-unresolved, gate-raised
≈mobx815.67 µs8.73×8.3%2102008/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈valtio816.62 µs8.74×2.0%2102128/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈baseline-array910.14 µs9.74×1.8%2101238/8 · 0/82026-10-22fine-unresolved, busy, gate-raised
#16tinybase1.841 ms20×4.1%2102148/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#17jotai2.241 ms24×5.6%2103058/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈oimdb-3.12.13.687 ms39×0.0%2102168/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#19legend-state4.378 ms47×3.3%2101828/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20tanstack-db4.886 ms52×2.0%2102498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed5.435 ms58×20.0%2102498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager5.467 ms59×1.2%2102498/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23zustand17.951 ms192×0.9%2102458/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#24redux-toolkit27.030 ms289×1.2%2102908/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25signaldb81.320 ms870×1.9%2102718/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26xstate-store579.126 ms6198×0.6%2102978/8 · 0/02026-10-22fine-unresolved, busy, gate-raised

a consumer per group consumerPerGroupnarrow

50 steps, with one view per group (50 views) instead of two. Each step reorders rows inside one group's page, so exactly one view may re-render. Measured: how many of the other 49 views were woken anyway. A library that notifies only the affected group wakes 50 views in total; one that notifies everybody wakes 50 × 50, and each of those views re-renders its page.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1preact-signals454.18 µs0.66×3.5%9029250/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈oimdb-sorted-index465.98 µs0.68×11.6%9027950/50 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#3alien-signals524.87 µs0.77×10.5%9032250/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈vue-reactivity528.01 µs0.77×2.5%9032650/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#5solid-js663.41 µs0.97×7.3%9031650/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈baseline-indexed685.90 µs1.00×2.3%9024050/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-store849.94 µs1.24×27.7%9023850/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈nanostores-per-row910.49 µs1.33×4.9%9021050/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈oimdb966.82 µs1.43×1.2%9027950/50 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#10effector1.192 ms1.74×6.1%9030850/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#11oimdb-3.12.11.706 ms2.49×6.9%9021050/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#12lokijs1.936 ms2.82×1.2%9020650/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#13nanostores2.128 ms3.10×5.4%9021050/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#14tanstack-db3.733 ms5.44×11.1%9024350/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager4.213 ms6.14×18.4%9024350/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed4.245 ms6.19×2.1%9024350/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#17valtio6.246 ms9.11×5.3%9020650/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#18mobx7.599 ms11×5.3%9019450/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tinybase8.032 ms12×15.6%9020850/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20jotai13.862 ms20×11.1%9029950/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#21legend-state32.252 ms47×1.7%9017650/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#22xstate-store142.197 ms207×1.3%9029150/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23zustand147.578 ms215×0.6%9023950/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#24baseline-array174.882 ms255×1.5%9011750/2500 · 0/02026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit199.892 ms291×1.0%9028450/50 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb1577.967 ms2301×1.5%9026550/50 · 0/02026-10-22fine-unresolved, busy, gate-raised

targeted column delivery cellDeliverynarrow

16 steps over 8 rows that are looked up once before timing starts. View A shows their `label`, view B their `note`. The first half of the steps changes each `label`, the second half each `note`. A change must reach only the view showing that field: a library that tracks fields wakes one view per change, one that tracks whole rows wakes both, and one that tracks nothing re-reads everything.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-indexed17.53 µs1.00×1.8%7,1402468/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈oimdb18.91 µs1.07×2.5%7,1402858/8 · 8/82026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈preact-signals18.69 µs1.08×4.0%7,1402988/8 · 8/82026-10-21fine-unresolved, gate-raised
≈alien-signals18.70 µs1.08×0.5%7,1403288/8 · 8/82026-10-21fine-unresolved, gate-raised
≈oimdb-sorted-index19.18 µs1.09×1.9%7,1402858/8 · 8/82026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈vue-reactivity20.49 µs1.18×0.4%7,1403328/8 · 8/82026-10-21fine-unresolved, gate-raised
≈tanstack-store20.93 µs1.19×1.4%7,1402448/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#8valtio23.82 µs1.36×1.6%7,1402128/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#9mobx27.13 µs1.55×1.3%7,1402008/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈solid-js27.00 µs1.55×2.1%7,1403228/8 · 8/82026-10-21fine-unresolved, gate-raised
#11baseline-array33.08 µs1.89×1.1%7,1401238/16 · 8/162026-10-22fine-unresolved, busy, gate-raised
#12lokijs35.61 µs2.03×0.7%7,1402128/16 · 8/162026-10-22fine-unresolved, busy, gate-raised
#13oimdb-3.12.138.38 µs2.19×0.9%7,1402168/16 · 8/162026-10-22fine-unresolved, busy, gate-raised
≈tinybase63.96 µs3.65×5.8%7,1402148/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈jotai65.08 µs3.71×3.8%7,1403058/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#16nanostores-per-row119.81 µs6.83×1.1%7,1402168/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#17legend-state130.82 µs7.46×3.2%7,1401828/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#18nanostores167.33 µs9.54×4.7%7,1402168/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#19effector247.61 µs14×4.7%7,1403148/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db517.18 µs29×7.8%7,1402498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager526.43 µs30×21.2%7,1402498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed543.31 µs31×3.4%7,1402498/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#23xstate-store36.082 ms2058×0.7%7,1402978/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
≈zustand36.236 ms2067×0.1%7,1402458/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#25redux-toolkit53.892 ms3074×0.9%7,1402908/8 · 8/82026-10-22fine-unresolved, busy, gate-raised
#26signaldb71.638 ms4086×2.4%7,1402718/16 · 8/162026-10-22fine-unresolved, busy, gate-raised

cell lower bound cellFloorsynthetic

Lower bound, no library: the steps of "targeted column delivery", each rendering one changed line of a 8-line view straight from an array of strings. A library that notifies only the view showing the changed field renders this much; one that notifies both views renders twice as much.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array16.47 µs3.4%23,0401238/8 · 8/82026-10-22gate-raised

delete the visible window deleteRowsnarrow

One step: all 32 rows view A shows, the top of group A, are deleted in one batch. View A must re-render once, refilled with the rows that were below it. View B shows a different group and must not change.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-indexed23.75 µs1.00×24.9%452461/1 · 0/02026-10-22busy, gate-raised
≈preact-signals34.25 µs1.44×7.8%452981/1 · 0/02026-10-22busy, gate-raised
≈vue-reactivity34.33 µs1.45×7.6%453321/1 · 0/02026-10-22busy, gate-raised
≈effector35.42 µs1.49×2.4%453141/1 · 0/02026-10-22busy, gate-raised
≈tanstack-store41.71 µs1.76×8.2%452441/1 · 0/02026-10-22busy, gate-raised
≈alien-signals43.88 µs1.85×5.7%453281/1 · 0/02026-10-22busy, gate-raised
≈oimdb-sorted-index51.50 µs2.17×18.5%452851/1 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores-per-row51.71 µs2.18×31.7%452161/1 · 0/02026-10-22busy, gate-raised
≈oimdb56.58 µs2.30×43.4%452851/1 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores59.50 µs2.51×6.7%452161/1 · 0/02026-10-22busy, gate-raised
≈solid-js66.79 µs2.81×5.7%453221/1 · 0/02026-10-22busy, gate-raised
≈valtio137.13 µs5.77×16.9%452121/1 · 0/02026-10-22busy, gate-raised
≈baseline-array154.92 µs6.52×3.7%451231/1 · 0/12026-10-22busy, gate-raised
≈mobx283.71 µs12×26.1%452001/1 · 0/02026-10-22busy, gate-raised
≈jotai463.13 µs20×7.4%453051/1 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.1513.71 µs22×12.8%452161/1 · 0/02026-10-22busy, gate-raised
≈tinybase785.63 µs33×11.2%452141/1 · 0/02026-10-22busy, gate-raised
≈legend-state1.099 ms46×3.9%451821/1 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager1.440 ms61×12.9%452491/1 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed1.537 ms65×23.4%452491/1 · 0/02026-10-22busy, gate-raised
≈tanstack-db1.664 ms70×10.6%452491/1 · 0/02026-10-22busy, gate-raised
#22xstate-store2.299 ms97×2.2%452971/1 · 0/02026-10-22busy, gate-raised
≈zustand2.324 ms98×1.3%452451/1 · 0/02026-10-22busy, gate-raised
#24lokijs3.900 ms164×0.6%452121/1 · 0/02026-10-22busy, gate-raised
#25redux-toolkit5.079 ms214×1.2%452901/1 · 0/02026-10-22busy, gate-raised
≈signaldb11.702 ms493×2.7%452711/1 · 0/02026-10-22busy, gate-raised

delete one visible row deleteOneRowrepresentative

8 steps. Each step deletes one row view A shows, from different positions on the page: js-framework-benchmark's "remove row". View A must re-render once per step and stay full, pulling up the next row from below. View B must not change. Compare with "delete the visible window", which deletes the whole page at once.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1preact-signals57.46 µs0.53×28.6%272988/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈alien-signals66.33 µs0.61×9.7%273288/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈vue-reactivity67.79 µs0.63×37.9%273328/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈oimdb-sorted-index81.46 µs0.74×8.2%272858/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈solid-js101.13 µs0.94×21.4%273228/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈baseline-indexed110.00 µs1.00×3.9%272468/8 · 0/02026-10-22busy, gate-raised
≈oimdb109.33 µs1.00×18.0%272858/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores-per-row116.71 µs1.06×5.5%272168/8 · 0/02026-10-22busy, gate-raised
#9tanstack-store144.75 µs1.32×6.2%272448/8 · 0/02026-10-22busy, gate-raised
≈effector168.38 µs1.53×3.1%273148/8 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.1307.46 µs2.80×8.1%272168/8 · 0/02026-10-22busy, gate-raised
#12nanostores531.08 µs4.83×1.6%272168/8 · 0/02026-10-22busy, gate-raised
≈valtio706.17 µs6.42×1.8%272128/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db775.63 µs7.05×6.7%272498/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager776.46 µs7.06×18.3%272498/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed838.08 µs7.62×6.0%272498/8 · 0/02026-10-22busy, gate-raised
#17baseline-array1.070 ms9.73×2.1%271238/8 · 0/82026-10-22busy, gate-raised
≈mobx1.063 ms9.84×14.2%272008/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈jotai1.258 ms11×11.0%273058/8 · 0/02026-10-22busy, gate-raised
≈tinybase1.278 ms12×43.7%272148/8 · 0/02026-10-22busy, gate-raised
≈legend-state4.885 ms44×5.5%271828/8 · 0/02026-10-22busy, gate-raised
#22lokijs10.916 ms99×1.4%272128/8 · 0/02026-10-22busy, gate-raised
#23xstate-store18.088 ms164×1.4%272978/8 · 0/02026-10-22busy, gate-raised
≈zustand18.123 ms165×1.1%272458/8 · 0/02026-10-22busy, gate-raised
≈redux-toolkit39.823 ms362×3.3%272908/8 · 0/02026-10-22busy, gate-raised
#26signaldb46.533 ms423×3.9%272718/8 · 0/02026-10-22busy, gate-raised

the group empties under the view drainGroupnarrow

Two steps, starting with group A empty. Step one inserts every row of group A, so its list is built from nothing; step two deletes all of them in one batch, so view A must end showing an empty list. No other scenario empties a list. Afterwards the store is checked for rows a partial delete left behind. View B shows a different group and must not change.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈effector78.35 µs0.96×10.4%3603142/2 · 0/02026-10-22busy, gate-raised
≈baseline-indexed81.97 µs1.00×7.6%3602462/2 · 0/02026-10-22busy, gate-raised
≈vue-reactivity95.16 µs1.16×16.1%3603322/2 · 0/02026-10-22busy, gate-raised
≈nanostores102.56 µs1.25×13.1%3602162/2 · 0/02026-10-22busy, gate-raised
≈tanstack-store114.94 µs1.40×12.7%3602442/2 · 0/02026-10-22busy, gate-raised
≈preact-signals116.86 µs1.43×4.3%3602982/2 · 0/02026-10-22busy, gate-raised
≈nanostores-per-row118.35 µs1.44×4.2%3602162/2 · 0/02026-10-22busy, gate-raised
≈alien-signals120.27 µs1.47×2.4%3603282/2 · 0/02026-10-22busy, gate-raised
≈oimdb145.16 µs1.69×0.6%3602852/2 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb-sorted-index142.24 µs1.74×6.0%3602852/2 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈solid-js149.47 µs1.82×16.4%3603222/2 · 0/02026-10-22busy, gate-raised
≈baseline-array276.02 µs3.37×1.4%3601232/2 · 0/22026-10-22busy, gate-raised
≈mobx668.61 µs8.16×9.7%3602002/2 · 0/02026-10-22busy, gate-raised
#14valtio1.100 ms13×0.9%3602122/2 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.11.230 ms15×4.9%3602162/2 · 0/02026-10-22busy, gate-raised
#16lokijs1.317 ms16×0.5%3602122/2 · 0/02026-10-22busy, gate-raised
#17jotai1.726 ms21×2.0%3603052/2 · 0/02026-10-22busy, gate-raised
#18legend-state3.638 ms44×7.7%3601822/2 · 0/02026-10-22busy, gate-raised
#19tinybase4.071 ms50×1.3%3602142/2 · 0/02026-10-22busy, gate-raised
#20zustand4.452 ms54×1.4%3602452/2 · 0/02026-10-22busy, gate-raised
≈xstate-store4.503 ms55×0.3%3602972/2 · 0/02026-10-22busy, gate-raised
≈tanstack-db4.592 ms56×6.1%3602492/2 · 0/02026-10-22busy, gate-raised
#23tanstack-db-eager5.359 ms65×6.1%3602492/2 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed5.424 ms66×2.2%3602492/2 · 0/02026-10-22busy, gate-raised
#25redux-toolkit8.543 ms104×1.1%3602902/2 · 0/02026-10-22busy, gate-raised
#26signaldb37.155 ms453×1.8%3602712/2 · 0/02026-10-22busy, gate-raised

re-send unchanged list resendUnchangedrepresentative

One step: the 32 rows view A shows are written again, as new objects, with exactly the values they already have, like a page fetched again from a server. The benchmark code does not filter them out, so this shows whether the library itself notices that nothing changed. No view may show new text; how often views were woken for nothing is counted.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈vue-reactivity1.69 µs0.14×3.8%7,6803320/0 · 0/02026-10-21fine-unresolved, gate-raised
≈preact-signals1.80 µs0.15×79.2%7,6802980/0 · 0/02026-10-21fine-unresolved, gate-raised
≈oimdb-3.12.12.60 µs0.22×9.2%7,6802160/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈solid-js3.10 µs0.27×41.9%7,6803220/0 · 0/02026-10-21fine-unresolved, gate-raised
≈oimdb3.15 µs0.27×2.2%7,6802850/0 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb-sorted-index3.23 µs0.27×8.2%7,6802850/0 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈alien-signals3.98 µs0.34×2.9%7,6803280/0 · 0/02026-10-21fine-unresolved, gate-raised
#8baseline-indexed11.83 µs1.00×1.0%7,6802460/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
#9mobx13.67 µs1.16×8.2%7,6802000/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#10nanostores-per-row17.47 µs1.48×2.0%7,6802160/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-store17.68 µs1.49×5.2%7,6802440/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
#12effector20.38 µs1.72×1.8%7,6803140/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈valtio20.68 µs1.75×0.8%7,6802120/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈nanostores20.77 µs1.76×2.8%7,6802160/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
#15lokijs23.24 µs1.96×2.9%7,6802120/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-indexed33.16 µs2.80×5.4%7,6802490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db34.19 µs2.89×3.7%7,6802490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈tanstack-db-eager34.97 µs2.96×3.5%7,6802490/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#19legend-state41.22 µs3.48×3.3%7,6801820/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#20tinybase80.56 µs6.81×2.6%7,6802140/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#21baseline-array114.90 µs9.71×2.4%7,6801230/1 · 0/12026-10-22fine-unresolved, busy, gate-raised
#22jotai140.14 µs12×1.9%7,6803050/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
#23redux-toolkit2.232 ms189×0.8%7,6802900/0 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈zustand2.243 ms190×1.4%7,6802450/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈xstate-store2.261 ms191×1.0%7,6802970/1 · 0/02026-10-22fine-unresolved, busy, gate-raised
#26signaldb8.830 ms746×7.4%7,6802710/0 · 0/02026-10-22fine-unresolved, busy, gate-raised

consumer lower bound consumerFloorsynthetic

Lower bound, no library: a page of 32 rows rendered 32 times straight from an array of plain objects. Every participant renders through the same code, so no library can render a page for less; one that stores rows as proxies or wrapped objects also pays to read through them.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array122.59 µs0.7%7,51512332/32 · 0/02026-10-22busy, gate-raised

mount lower bound mountFloorsynthetic

Lower bound, no library: 8 cycles of subscribing 256 callbacks to one plain object, changing it, running them, unsubscribing all, changing it again. Subscribers live in a plain array with constant-time removal. No library can subscribe and unsubscribe for less; the mount and unmount tables are scaled by it.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array449.27 µs0.3%2,0251238/8 · 0/02026-10-22gate-raised

mount and unmount one subscriber churnOnenarrow

8 cycles of: subscribe one component to one row, change the row; unsubscribe, change it again. The smallest mount and unmount, which every other mount and unmount scenario is compared with. The first change must reach the component once; the second must reach nobody.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index2.58 µs0.94×5.2%7,6801558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb2.70 µs0.95×86.6%7,6801558/8 · 0/02026-10-31stale: adapter, stale: subject, unresolved, busy, gate-raised
≈baseline-array2.84 µs1.00×5.1%7,6801518/8 · 0/02026-10-31unresolved, busy, gate-raised
≈nanostores-per-row2.91 µs1.06×0.8%7,6801048/8 · 0/02026-10-22busy, gate-raised
≈nanostores2.96 µs1.08×2.1%7,6801048/8 · 0/02026-10-22busy, gate-raised
≈alien-signals3.04 µs1.13×5.6%7,6801418/8 · 0/02026-10-21gate-raised
≈tanstack-store3.10 µs1.13×6.4%7,6801308/8 · 0/02026-10-22busy, gate-raised
≈lokijs4.48 µs1.63×0.7%7,6801728/8 · 0/02026-10-22busy, gate-raised
≈preact-signals4.44 µs1.65×15.0%7,6801438/8 · 0/02026-10-21gate-raised
≈oimdb-3.12.16.68 µs2.35×445.7%7,6801178/8 · 0/02026-10-31unresolved, busy, gate-raised
≈vue-reactivity6.87 µs2.55×3.1%7,6801728/8 · 0/02026-10-21gate-raised
≈valtio8.22 µs2.99×0.7%7,6801168/8 · 0/02026-10-22busy, gate-raised
≈solid-js8.16 µs3.02×5.3%7,6801408/8 · 0/02026-10-21gate-raised
≈jotai13.34 µs4.86×2.2%7,6801628/8 · 0/02026-10-22busy, gate-raised
≈effector14.84 µs5.40×13.2%7,6801798/8 · 0/02026-10-22busy, gate-raised
≈mobx15.67 µs5.70×8.1%7,6801528/8 · 0/02026-10-22busy, gate-raised
#17tinybase35.94 µs13×5.8%7,680958/8 · 0/02026-10-22busy, gate-raised
≈legend-state48.75 µs18×0.3%7,6801148/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db645.46 µs235×5.9%7,6801898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed658.21 µs240×10.2%7,6801898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager682.50 µs248×0.7%7,6801898/8 · 0/02026-10-22busy, gate-raised
#22signaldb30.881 ms11242×3.2%7,6802108/8 · 0/02026-10-22busy, gate-raised
#23zustand35.767 ms13021×0.8%7,6802158/8 · 0/02026-10-22busy, gate-raised
≈xstate-store36.014 ms13111×0.6%7,6802288/8 · 0/02026-10-22busy, gate-raised
#25redux-toolkit53.751 ms19569×0.2%7,6802258/8 · 0/02026-10-22busy, gate-raised
Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N subscribers on one root, oldest out first churnSharednarrow

8 cycles of: subscribe 32 components to the same row, change it, unsubscribe them oldest first, change it again. This is a list of components bound to one shared object being mounted and unmounted. Compare with "mount and unmount one subscriber" for the cost per component, and with the N×8 variant for whether that cost grows with the number already subscribed.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array18.77 µs1.00×8.7%2,2351518/8 · 0/02026-10-22busy, gate-raised
≈oimdb20.39 µs1.01×1.4%2,2351558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores19.47 µs1.04×4.0%2,2351048/8 · 0/02026-10-22busy, gate-raised
≈nanostores-per-row19.57 µs1.04×0.8%2,2351048/8 · 0/02026-10-22busy, gate-raised
≈oimdb-sorted-index20.41 µs1.09×1.8%2,2351558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
#6tanstack-store32.30 µs1.72×5.8%2,2351308/8 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.137.33 µs1.99×9.8%2,2351178/8 · 0/02026-10-22busy, gate-raised
≈alien-signals40.71 µs2.17×1.4%2,2351418/8 · 0/02026-10-22busy, gate-raised
#9lokijs43.17 µs2.30×2.4%2,2351728/8 · 0/02026-10-22busy, gate-raised
≈preact-signals62.77 µs3.34×64.9%2,2351438/8 · 0/02026-10-22busy, gate-raised
≈valtio89.16 µs4.75×1.3%2,2351168/8 · 0/02026-10-22busy, gate-raised
≈jotai95.41 µs5.08×0.6%2,2351628/8 · 0/02026-10-22busy, gate-raised
≈solid-js130.38 µs6.94×24.5%2,2351408/8 · 0/02026-10-22busy, gate-raised
#14tinybase163.14 µs8.69×3.7%2,235958/8 · 0/02026-10-22busy, gate-raised
#15vue-reactivity177.12 µs9.43×5.3%2,2351728/8 · 0/02026-10-22busy, gate-raised
#16effector238.41 µs13×4.9%2,2351798/8 · 0/02026-10-22busy, gate-raised
#17mobx257.37 µs14×6.1%2,2351528/8 · 0/02026-10-22busy, gate-raised
≈legend-state464.61 µs25×0.8%2,2351148/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db1.183 ms63×10.0%2,2351898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed1.186 ms63×4.7%2,2351898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager1.280 ms68×19.7%2,2351898/8 · 0/02026-10-22busy, gate-raised
#22zustand35.726 ms1903×0.7%2,2352158/8 · 0/02026-10-22busy, gate-raised
≈xstate-store35.797 ms1907×0.5%2,2352288/8 · 0/02026-10-22busy, gate-raised
#24signaldb44.707 ms2381×2.3%2,2352108/8 · 0/02026-10-22busy, gate-raised
#25redux-toolkit53.063 ms2826×1.3%2,2352258/8 · 0/02026-10-22busy, gate-raised
Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N subscribers on one root, newest out first churnSharedReversednarrow

The same as "N subscribers on one root, oldest out first", but the 32 components unsubscribe newest first. The difference shows whether the order of unmounting matters: a library that keeps subscribers in an array pays differently for removing from its front and from its back.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index20.21 µs0.99×2.0%2,1601558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈baseline-array20.43 µs1.00×6.8%2,1601518/8 · 0/02026-10-22busy, gate-raised
≈oimdb20.12 µs1.00×5.1%2,1601558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈nanostores-per-row21.27 µs1.04×0.7%2,1601048/8 · 0/02026-10-22busy, gate-raised
≈nanostores21.46 µs1.05×1.6%2,1601048/8 · 0/02026-10-22busy, gate-raised
#6lokijs32.61 µs1.60×2.6%2,1601728/8 · 0/02026-10-22busy, gate-raised
≈tanstack-store33.18 µs1.62×2.7%2,1601308/8 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.136.87 µs1.80×8.7%2,1601178/8 · 0/02026-10-22busy, gate-raised
#9alien-signals40.58 µs1.99×1.8%2,1601418/8 · 0/02026-10-22busy, gate-raised
#10valtio88.12 µs4.31×0.2%2,1601168/8 · 0/02026-10-22busy, gate-raised
#11jotai94.08 µs4.60×2.0%2,1601628/8 · 0/02026-10-22busy, gate-raised
#12preact-signals103.85 µs5.08×4.2%2,1601438/8 · 0/02026-10-22busy, gate-raised
≈solid-js121.30 µs5.94×36.8%2,1601408/8 · 0/02026-10-22busy, gate-raised
≈vue-reactivity139.51 µs6.83×30.8%2,1601728/8 · 0/02026-10-22busy, gate-raised
≈tinybase161.30 µs7.89×2.6%2,160958/8 · 0/02026-10-22busy, gate-raised
#16effector250.67 µs12×3.7%2,1601798/8 · 0/02026-10-22busy, gate-raised
≈mobx257.87 µs13×3.6%2,1601528/8 · 0/02026-10-22busy, gate-raised
#18legend-state463.00 µs23×0.6%2,1601148/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-indexed1.054 ms52×9.1%2,1601898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db1.090 ms53×6.0%2,1601898/8 · 0/02026-10-22busy, gate-raised
≈tanstack-db-eager1.215 ms59×14.5%2,1601898/8 · 0/02026-10-22busy, gate-raised
#22zustand35.644 ms1744×0.3%2,1602158/8 · 0/02026-10-22busy, gate-raised
≈xstate-store35.744 ms1749×0.3%2,1602288/8 · 0/02026-10-22busy, gate-raised
#24signaldb45.798 ms2241×0.4%2,1602108/8 · 0/02026-10-22busy, gate-raised
#25redux-toolkit53.324 ms2610×0.9%2,1602258/8 · 0/02026-10-22busy, gate-raised
Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N×8 subscribers on one root, oldest out first churnSharedDeepsynthetic

The same as "N subscribers on one root, oldest out first" with 256 components instead of 32. If unsubscribing is linear, eight times the components cost eight times more; sixty-four times more means it is quadratic. One size cannot tell those apart, which is why there are two.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb147.95 µs0.85×1.6%3001558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb-sorted-index147.19 µs0.88×1.0%3001558/8 · 0/02026-10-23stale: adapter, stale: subject, busy, gate-raised
#3baseline-array166.70 µs1.00×0.8%3001518/8 · 0/02026-10-23busy, gate-raised
#4nanostores178.23 µs1.07×3.6%3001048/8 · 0/02026-10-23busy, gate-raised
≈nanostores-per-row179.08 µs1.07×2.4%3001048/8 · 0/02026-10-23busy, gate-raised
#6oimdb-3.12.1276.91 µs1.66×6.8%3001178/8 · 0/02026-10-23busy, gate-raised
≈tanstack-store286.19 µs1.72×3.0%3001308/8 · 0/02026-10-23busy, gate-raised
#8alien-signals319.58 µs1.92×7.6%3001418/8 · 0/02026-10-23busy, gate-raised
#9jotai671.17 µs4.03×0.8%3001628/8 · 0/02026-10-23busy, gate-raised
≈valtio681.31 µs4.09×3.1%3001168/8 · 0/02026-10-23busy, gate-raised
#11preact-signals834.70 µs5.01×11.7%3001438/8 · 0/02026-10-23busy, gate-raised
≈solid-js979.96 µs5.88×32.0%3001408/8 · 0/02026-10-23busy, gate-raised
≈lokijs1.013 ms6.08×0.7%3001728/8 · 0/02026-10-23busy, gate-raised
≈tinybase1.106 ms6.63×13.5%300958/8 · 0/02026-10-23busy, gate-raised
≈vue-reactivity1.131 ms6.78×2.7%3001728/8 · 0/02026-10-23busy, gate-raised
≈effector1.711 ms10×4.1%3001798/8 · 0/02026-10-23busy, gate-raised
≈mobx1.748 ms10×2.2%3001528/8 · 0/02026-10-23busy, gate-raised
#18legend-state3.411 ms20×1.9%3001148/8 · 0/02026-10-23busy, gate-raised
#19tanstack-db-indexed7.268 ms44×2.4%3001898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db7.354 ms44×13.5%3001898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db-eager7.952 ms48×5.8%3001898/8 · 0/02026-10-23busy, gate-raised
#22zustand35.708 ms214×0.8%3002158/8 · 0/02026-10-23busy, gate-raised
#23xstate-store36.101 ms217×0.9%3002288/8 · 0/02026-10-23busy, gate-raised
#24redux-toolkit53.959 ms324×1.5%3002258/8 · 0/02026-10-23busy, gate-raised

✗ signaldb — failed in time: spawnSync /Users/abaikov/.nvm/versions/node/v24.16.0/bin/node ENOBUFS

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N×8 subscribers on one root, newest out first churnSharedDeepReversedsynthetic

256 components on one row, unsubscribing newest first. With the other three variants it shows whether the effect of unmount order grows with the number of components or stays constant.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index147.70 µs0.70×1.2%3001558/8 · 0/02026-10-23stale: adapter, stale: subject, busy, gate-raised
≈oimdb146.71 µs0.73×13.8%3001558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
#3nanostores-per-row205.94 µs0.98×11.7%3001048/8 · 0/02026-10-23busy, gate-raised
≈baseline-array210.29 µs1.00×3.8%3001518/8 · 0/02026-10-23busy, gate-raised
≈nanostores217.57 µs1.03×1.9%3001048/8 · 0/02026-10-23busy, gate-raised
≈oimdb-3.12.1252.00 µs1.20×8.5%3001178/8 · 0/02026-10-23busy, gate-raised
#7tanstack-store273.97 µs1.30×0.4%3001308/8 · 0/02026-10-23busy, gate-raised
#8alien-signals328.99 µs1.56×2.6%3001418/8 · 0/02026-10-23busy, gate-raised
≈lokijs335.91 µs1.60×5.2%3001728/8 · 0/02026-10-23busy, gate-raised
≈jotai652.99 µs3.11×0.3%3001628/8 · 0/02026-10-23busy, gate-raised
#11valtio670.05 µs3.19×1.3%3001168/8 · 0/02026-10-23busy, gate-raised
#12preact-signals794.25 µs3.78×2.8%3001438/8 · 0/02026-10-23busy, gate-raised
#13tinybase1.093 ms5.20×6.5%300958/8 · 0/02026-10-23busy, gate-raised
≈vue-reactivity1.147 ms5.46×0.6%3001728/8 · 0/02026-10-23busy, gate-raised
≈solid-js1.292 ms6.14×5.8%3001408/8 · 0/02026-10-23busy, gate-raised
≈mobx1.698 ms8.07×32.8%3001528/8 · 0/02026-10-23busy, gate-raised
≈effector1.827 ms8.69×2.9%3001798/8 · 0/02026-10-23busy, gate-raised
#18legend-state3.439 ms16×3.6%3001148/8 · 0/02026-10-23busy, gate-raised
#19tanstack-db6.803 ms32×16.1%3001898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db-indexed6.856 ms33×2.5%3001898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db-eager7.255 ms35×2.9%3001898/8 · 0/02026-10-23busy, gate-raised
#22zustand35.862 ms171×0.2%3002158/8 · 0/02026-10-23busy, gate-raised
#23xstate-store36.292 ms173×0.8%3002288/8 · 0/02026-10-23busy, gate-raised
#24redux-toolkit53.917 ms256×0.0%3002258/8 · 0/02026-10-23busy, gate-raised

✗ signaldb — failed in time: spawnSync /Users/abaikov/.nvm/versions/node/v24.16.0/bin/node ENOBUFS

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

one subscriber over many sources churnWidenarrow

8 cycles of: subscribe one component that reads 64 rows, change the first of them, unsubscribe, change it again. The other variants vary how many components watch one row; this one varies how many rows one component watches, which is what a subscription pays to register with each of its sources and to leave them. Compare with "mount and unmount one subscriber", whose component reads one row.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index13.41 µs0.53×3.5%2,3851558/8 · 0/02026-10-23stale: adapter, stale: subject, busy, gate-raised
≈oimdb13.82 µs0.54×33.7%2,3851558/8 · 0/02026-10-31busy, gate-raised
≈tanstack-store17.52 µs0.69×7.9%2,3851308/8 · 0/02026-10-23busy, gate-raised
≈baseline-array25.49 µs1.00×19.0%2,3851518/8 · 0/02026-10-31busy, gate-raised
≈lokijs26.13 µs1.04×1.8%2,3851728/8 · 0/02026-10-23busy, gate-raised
#6valtio61.14 µs2.42×2.4%2,3851168/8 · 0/02026-10-23busy, gate-raised
≈alien-signals64.05 µs2.64×0.7%2,3851418/8 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.168.11 µs2.70×3.4%2,3851178/8 · 0/02026-10-23busy, gate-raised
≈preact-signals102.15 µs4.21×0.5%2,3851438/8 · 0/02026-10-22busy, gate-raised
#10vue-reactivity127.34 µs5.24×6.3%2,3851728/8 · 0/02026-10-22busy, gate-raised
#11tinybase273.55 µs11×4.1%2,385958/8 · 0/02026-10-23busy, gate-raised
≈solid-js264.71 µs11×3.1%2,3851408/8 · 0/02026-10-22busy, gate-raised
#13mobx328.90 µs14×6.8%2,3851528/8 · 0/02026-10-22busy, gate-raised
≈legend-state358.14 µs14×2.3%2,3851148/8 · 0/02026-10-23busy, gate-raised
≈jotai373.74 µs15×4.8%2,3851628/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db-indexed501.37 µs20×0.8%2,3851898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db504.72 µs20×115.1%2,3851898/8 · 0/02026-10-23busy, gate-raised
≈tanstack-db-eager517.41 µs21×15.5%2,3851898/8 · 0/02026-10-23busy, gate-raised
≈effector671.87 µs27×3.0%2,3851798/8 · 0/02026-10-23busy, gate-raised
#20signaldb32.484 ms1288×1.5%2,3852108/8 · 0/02026-10-23busy, gate-raised
#21xstate-store35.569 ms1410×1.3%2,3852288/8 · 0/02026-10-23busy, gate-raised
≈zustand35.616 ms1412×1.1%2,3852158/8 · 0/02026-10-23busy, gate-raised
≈nanostores42.787 ms1696×2.7%2,3851048/8 · 0/02026-10-23busy, gate-raised
≈nanostores-per-row42.873 ms1699×0.5%2,3851048/8 · 0/02026-10-23busy, gate-raised
#25redux-toolkit53.222 ms2110×0.8%2,3852258/8 · 0/02026-10-23busy, gate-raised
Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N subscribers on one column churnColumnnarrow

The same as "N subscribers on one root, oldest out first" (32 components, 8 cycles), but each component subscribes to one field of the row, `label`, instead of the whole row. The difference is what subscribing to a single field costs. A library with no per-field subscription skips this scenario rather than repeat its whole-row number.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array18.89 µs1.00×10.0%2,3701488/8 · 0/02026-10-23busy, gate-raised
≈alien-signals20.32 µs1.05×1.5%2,3701318/8 · 0/02026-10-22busy, gate-raised
≈oimdb21.93 µs1.12×2.8%2,3701578/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb-sorted-index21.96 µs1.16×0.8%2,3701578/8 · 0/02026-10-23stale: adapter, stale: subject, busy, gate-raised
≈preact-signals29.94 µs1.55×3.0%2,3701338/8 · 0/02026-10-22busy, gate-raised
≈nanostores-per-row29.74 µs1.57×12.3%2,370918/8 · 0/02026-10-23busy, gate-raised
≈nanostores30.10 µs1.59×4.0%2,370918/8 · 0/02026-10-23busy, gate-raised
#8oimdb-3.12.140.34 µs2.14×20.6%2,3701018/8 · 0/02026-10-23busy, gate-raised
#9valtio50.69 µs2.68×2.0%2,3701038/8 · 0/02026-10-23busy, gate-raised
≈tanstack-store54.67 µs2.89×15.7%2,3701278/8 · 0/02026-10-23busy, gate-raised
≈solid-js72.50 µs3.76×3.3%2,3701408/8 · 0/02026-10-22busy, gate-raised
≈vue-reactivity97.43 µs5.05×43.8%2,3701728/8 · 0/02026-10-22busy, gate-raised
#13mobx168.82 µs8.74×1.5%2,3701438/8 · 0/02026-10-22busy, gate-raised
≈tinybase167.75 µs8.88×6.5%2,370898/8 · 0/02026-10-23busy, gate-raised
#15legend-state574.35 µs30×1.2%2,3701058/8 · 0/02026-10-23busy, gate-raised
#16jotai729.21 µs39×1.8%2,3701518/8 · 0/02026-10-23busy, gate-raised
#17effector1.489 ms79×9.0%2,3701698/8 · 0/02026-10-23busy, gate-raised
#18zustand35.631 ms1886×0.1%2,3702108/8 · 0/02026-10-23busy, gate-raised
#19xstate-store35.802 ms1895×0.3%2,3702158/8 · 0/02026-10-23busy, gate-raised
#20signaldb45.924 ms2431×2.3%2,3702088/8 · 0/02026-10-23busy, gate-raised
#21redux-toolkit53.424 ms2828×0.8%2,3702208/8 · 0/02026-10-23busy, gate-raised

✗ tanstack-db — failed in time: spawnSync /Users/abaikov/.nvm/versions/node/v24.16.0/bin/node ETIMEDOUT

✗ tanstack-db-eager — failed in time: Command failed: /Users/abaikov/.nvm/versions/node/v24.16.0/bin/node --expose-gc --import tsx /Users/abaikov/work/oimdb/packages/bench/src/worker.ts tanstack-db-eager churnColumn 10000 time 158 5 50 / / <--- Last few GCs --->

✗ tanstack-db-indexed — failed in time: Command failed: /Users/abaikov/.nvm/versions/node/v24.16.0/bin/node --expose-gc --import tsx /Users/abaikov/work/oimdb/packages/bench/src/worker.ts tanstack-db-indexed churnColumn 10000 time 158 5 50 / / <--- Last few GCs --->

Declined by baseline-indexed, lokijs

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

lokijs: LokiJS HAS NO COLUMN IDENTITY. `update` is document-level: the event carries the document, and with cloning off the "old" document it also carries is THE SAME OBJECT, so nothing in the library can say which column moved. Telling them apart needs either a cloned collection with the adapter comparing old and new labels, or the adapter remembering the last label, and both are the adapter doing the library's job, which the B3 clause of churnColumn forbids. The row and wide shapes are supplied and unaffected.

N subscribers on N rows, a list churnListrepresentative

8 cycles of: subscribe 32 components, each to its own row, the way a list mounts one component per row; change some of those rows in one batch; unsubscribe them oldest first; change the same rows again. Each cycle changes 4 fewer rows, so the number of components that re-render proves each watches its own row. After unsubscribing nobody may be notified. Compare with "N subscribers on one root, oldest out first", where all components share one row.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index26.88 µs0.91×10.6%1,1101558/8 · 0/02026-10-23stale: adapter, stale: subject, busy, gate-raised
≈oimdb27.60 µs0.91×2.7%1,1101558/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈alien-signals29.37 µs0.96×2.7%1,1101418/8 · 0/02026-10-18stale: harness, stale: adapter, busy, gate-raised
≈baseline-array29.52 µs1.00×11.6%1,1101518/8 · 0/02026-10-23busy, gate-raised
≈preact-signals49.78 µs1.63×8.3%1,1101438/8 · 0/02026-10-18stale: harness, stale: adapter, busy, gate-raised
#6oimdb-3.12.166.10 µs2.24×3.1%1,1101178/8 · 0/02026-10-23busy, gate-raised
#7tinybase424.68 µs14×5.1%1,110958/8 · 0/02026-10-23busy, gate-raised

Not yet measured: signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

N subscribers over a derivation chain churnChainnarrow

A dashboard of 32 tiles; tile i summarises rows i to i+7 through a chain of 4 derived values, built when the tile mounts, like a component's selectors. 8 cycles of: mount all tiles, change row 0, unmount them oldest first, change row 0 again. Only tile 0 reads row 0, so exactly one tile may update, and after unmounting none may. Measured: building chains of derived values on mount and taking them apart on unmount.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈alien-signals153.48 µs0.60×85.0%601428/8 · 0/02026-10-21unattested, gate-raised
≈baseline-array260.59 µs1.00×1.5%601678/8 · 0/02026-10-23unattested, gate-raised
#3oimdb-sorted-index303.43 µs1.16×2.9%602298/8 · 0/02026-10-23stale: adapter, stale: subject, unattested, gate-raised
≈oimdb309.22 µs1.17×4.6%602298/8 · 0/02026-10-22stale: adapter, stale: subject, unattested, gate-raised
#5solid-js390.72 µs1.52×10.5%601478/8 · 0/02026-10-21unattested, gate-raised
≈vue-reactivity409.96 µs1.59×7.8%601788/8 · 0/02026-10-21unattested, gate-raised
≈preact-signals880.79 µs3.43×24.2%601448/8 · 0/02026-10-21unattested, gate-raised
≈mobx1.973 ms7.67×5.7%601548/8 · 0/02026-10-18stale: harness, stale: adapter, unattested, busy, gate-raised
≈jotai2.960 ms12×3.3%601588/8 · 0/02026-10-18stale: harness, stale: adapter, unattested, busy, gate-raised
#10redux-toolkit53.423 ms208×3.3%602248/8 · 0/02026-10-18stale: harness, stale: adapter, stale: library, unattested, busy, gate-raised

✗ oimdb-3.12.1 — failed in verify: the last churn line is "rd4yv000 1 32:sxxxxxxxxxxxxxx,lqwpk001,l4og9002,lig6y003,lw7xn004,l9zoc005,lnrf1006,l1j5q007>1>2>3>4#32/91809892", and the script's last delivering frame says "rd4yv000 1 1:sxxxxxxxxxxxxxx,lqwpk001,l4og9002,lig6y003,lw7xn004,l9zoc005,lnrf1006,l1j5q007>1>2>3>4#32/91809892"

✗ tinybase — failed in verify: the last churn line is "rd4yv000 1 1:sxxxxxxxxxxxxxx>1>2>3>4#32/ad44e32", and the script's last delivering frame says "rd4yv000 1 1:sxxxxxxxxxxxxxx,lqwpk001,l4og9002,lig6y003,lw7xn004,l9zoc005,lnrf1006,l1j5q007>1>2>3>4#32/91809892"

Not yet measured: signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, legend-state, effector, nanostores, nanostores-per-row, zustand, xstate-store, tanstack-store.

Declined by baseline-indexed, lokijs, valtio

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

lokijs: LokiJS HAS NO DERIVED VALUE A SECOND DERIVATION CAN READ. A DynamicView is a filtered, sorted list over one collection and cannot take another view as its source, and the collection emitter has no per-source channel, so a mount would be the wide shape's collection listener with the depth levels re-run by hand inside its body, which the B3 clause of churnChain forbids: the adapter building the chain, not the library, and the wide shape's row printed under the chain's name. The row and wide shapes are supplied and unaffected.

valtio: VALTIO 2.3.2 HAS NO DERIVED VALUE A SECOND DERIVATION CAN READ. Its only primitives are `proxy`, `subscribe` and `subscribeKey`; the derived-proxy helper lives in the separate `derive-valtio` package, which this suite does not install. A chain would be plain proxies that this adapter rewrites level by level inside its own subscribe callbacks, which the B3 clause of churnChain forbids: the adapter building the chain, not the library. The row, wide and column shapes are supplied and unaffected.

a write through standing chains chainFrameWritenarrow

The same 32-tile dashboard as "N subscribers over a derivation chain", mounted before timing starts. 16 steps, each changing row 7: the 8 tiles that read it must recompute their chains and the rest must not wake. Nothing is mounted or unmounted, so this is the cost of running the chains; the other scenario is the cost of building them.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array27.73 µs1.00×1.9%30016716/16 · 0/02026-10-23unattested, gate-raised
#2oimdb-sorted-index51.74 µs1.87×1.7%30022916/16 · 0/02026-10-23stale: adapter, stale: subject, unattested, gate-raised
≈oimdb51.71 µs1.87×1.2%30022916/16 · 0/02026-10-22stale: adapter, stale: subject, unattested, gate-raised
#4alien-signals56.83 µs2.06×3.1%30014216/16 · 0/02026-10-21unattested, gate-raised
≈preact-signals60.15 µs2.18×3.8%30014416/16 · 0/02026-10-21unattested, gate-raised
≈vue-reactivity74.53 µs2.70×1.6%30017816/16 · 0/02026-10-21unattested, gate-raised
#7oimdb-3.12.1111.33 µs4.02×4.0%30013316/16 · 0/02026-10-23unattested, gate-raised
#8solid-js129.70 µs4.70×13.2%30014716/16 · 0/02026-10-21unattested, gate-raised
#9mobx367.44 µs13×6.2%30015416/16 · 0/02026-10-18stale: harness, stale: adapter, unattested, gate-raised
#10jotai1.118 ms40×1.2%30015816/16 · 0/02026-10-18stale: harness, stale: adapter, unattested, gate-raised
#11redux-toolkit54.568 ms1973×1.0%30022416/16 · 0/02026-10-18stale: harness, stale: adapter, stale: library, unattested, gate-raised

✗ tinybase — failed in verify: the last churn line is "rd4yv000 1 1:ld4yv000>1>2>3>4#32/a0e0819", and the script's last delivering frame says "rd4yv000 1 8:ld4yv000,lqwpk001,l4og9002,lig6y003,lw7xn004,l9zoc005,lnrf1006,sxxxxxxxxxxxxxxx>1>2>3>4#32/f1b8dada"

Not yet measured: signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, legend-state, effector, nanostores, nanostores-per-row, zustand, xstate-store, tanstack-store.

Declined by baseline-indexed, lokijs, valtio

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

lokijs: LokiJS HAS NO DERIVED VALUE A SECOND DERIVATION CAN READ. A DynamicView is a filtered, sorted list over one collection and cannot take another view as its source, and the collection emitter has no per-source channel, so a mount would be the wide shape's collection listener with the depth levels re-run by hand inside its body, which the B3 clause of churnChain forbids: the adapter building the chain, not the library, and the wide shape's row printed under the chain's name. The row and wide shapes are supplied and unaffected.

valtio: VALTIO 2.3.2 HAS NO DERIVED VALUE A SECOND DERIVATION CAN READ. Its only primitives are `proxy`, `subscribe` and `subscribeKey`; the derived-proxy helper lives in the separate `derive-valtio` package, which this suite does not install. A chain would be plain proxies that this adapter rewrites level by level inside its own subscribe callbacks, which the B3 clause of churnChain forbids: the adapter building the chain, not the library. The row, wide and column shapes are supplied and unaffected.

N subscribers over a diamond churnDiamondnarrow

8 cycles of: mount 32 components, each building its own diamond, two values derived from row 0 and a third that combines both; change row 0; unmount them oldest first; change it again. The combined value must be computed once per change, not once per branch: computing it twice is the "glitch" this checks for. Compare with "N subscribers over a derivation chain", where the derived values form a line.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈alien-signals110.04 µs0.85×11.3%1051358/8 · 0/02026-10-21unresolved, gate-raised
#2baseline-array129.41 µs1.00×2.2%1051618/8 · 0/02026-10-23gate-raised
#3oimdb201.15 µs1.52×7.5%1052298/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb-sorted-index203.89 µs1.58×0.9%1052298/8 · 0/02026-10-23stale: adapter, stale: subject, gate-raised
≈preact-signals286.50 µs2.22×38.9%1051378/8 · 0/02026-10-21unresolved, gate-raised
≈vue-reactivity307.08 µs2.38×14.8%1051768/8 · 0/02026-10-21unresolved, gate-raised
≈solid-js407.04 µs3.16×5.9%1051458/8 · 0/02026-10-21unresolved, gate-raised
≈oimdb-3.12.11.201 ms9.28×17.6%1051238/8 · 0/02026-10-23gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, xstate-store, tanstack-store.

Declined by baseline-indexed, lokijs, valtio

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

lokijs: LOKIJS HAS NO DERIVATION PRIMITIVE. Its derived structures are DynamicViews and Resultsets, filtered and sorted lists over ONE collection that cannot take a view as a source, and nothing in it computes a value from two others; the two branches and the join could only be functions this adapter calls by hand inside a collection listener, which the B3 clause of churnDiamond forbids: the adapter building the diamond, not the library. The row and wide shapes are supplied and unaffected.

valtio: VALTIO 2.3.2 HAS NO DERIVATION PRIMITIVE. Nothing in `valtio/vanilla` computes a value from two others, so the two branches and the join could only be proxies this adapter rewrites by hand, and since the join is rewritten once per branch, delivering it once needs the adapter to hold the wake back and dedup it in settle, which the B3 clause of churnDiamond forbids: the adapter building the diamond and its glitch-freedom, not the library. The row, wide and column shapes are supplied and unaffected.

a write through standing diamonds diamondFrameWrite

The same 32 diamonds as "N subscribers over a diamond", mounted before timing starts. 16 steps, each changing row 0: every diamond recomputes once per step. Nothing is mounted or unmounted, so this is the cost of running diamonds; the other scenario is the cost of building them.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array21.37 µs1.00×0.6%93016116/16 · 0/02026-10-23unresolved, gate-raised
#2alien-signals78.31 µs3.66×1.3%93013516/16 · 0/02026-10-21unresolved, busy, gate-raised
≈oimdb-sorted-index78.34 µs3.67×4.0%93022916/16 · 0/02026-10-23stale: adapter, stale: subject, unresolved, gate-raised
≈oimdb79.16 µs3.73×0.7%93022916/16 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈preact-signals90.85 µs4.24×4.9%93013716/16 · 0/02026-10-21unresolved, busy, gate-raised
#6vue-reactivity141.67 µs6.62×3.8%93017616/16 · 0/02026-10-21unresolved, busy, gate-raised
#7solid-js282.82 µs13×5.6%93014516/16 · 0/02026-10-21unresolved, busy, gate-raised
≈oimdb-3.12.1332.13 µs16×2.1%93012316/16 · 0/02026-10-23unresolved, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, xstate-store, tanstack-store.

Declined by baseline-indexed, lokijs, valtio

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

lokijs: LOKIJS HAS NO DERIVATION PRIMITIVE. Its derived structures are DynamicViews and Resultsets, filtered and sorted lists over ONE collection that cannot take a view as a source, and nothing in it computes a value from two others; the two branches and the join could only be functions this adapter calls by hand inside a collection listener, which the B3 clause of churnDiamond forbids: the adapter building the diamond, not the library. The row and wide shapes are supplied and unaffected.

valtio: VALTIO 2.3.2 HAS NO DERIVATION PRIMITIVE. Nothing in `valtio/vanilla` computes a value from two others, so the two branches and the join could only be proxies this adapter rewrites by hand, and since the join is rewritten once per branch, delivering it once needs the adapter to hold the wake back and dedup it in settle, which the B3 clause of churnDiamond forbids: the adapter building the diamond and its glitch-freedom, not the library. The row, wide and column shapes are supplied and unaffected.

churn against a standing population churnResidentnarrow

256 components stay subscribed to one row the whole time. 8 cycles of: subscribe 32 more to the same row, change it, unsubscribe those, change it again. A real screen keeps most of its subscriptions and mounts or unmounts a few; this shows whether that costs more when many subscribers are already there.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array67.13 µs1.00×1.3%63015116/16 · 0/02026-10-23gate-raised
#2oimdb-sorted-index93.90 µs1.40×4.5%63015516/16 · 0/02026-10-23stale: adapter, stale: subject, gate-raised
≈oimdb93.93 µs1.41×10.1%63015516/16 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈alien-signals101.99 µs1.52×1.1%63014116/16 · 0/02026-10-18stale: harness, stale: adapter, busy, gate-raised
#5preact-signals148.68 µs2.22×3.4%63014316/16 · 0/02026-10-18stale: harness, stale: adapter, busy, gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

an application screen fills appFillnarrow

An app screen: 80 cards in 10 decks owned by 5 users, already in the store. In one step, 180 components mount (one per card, one per tag shown on a card, one per deck name, and one per deck date, which reads a setting of the deck's user) and 46 of the 95 records change. The only scenario where data changes and components mount together. Exactly 87 components, those whose own record changed, must re-render.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈oimdb-sorted-index23.13 µs0.93×12.8%271551/1 · 0/02026-10-23stale: adapter, stale: subject, gate-raised
≈oimdb22.08 µs0.95×6.6%271551/1 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈baseline-array24.92 µs1.00×10.2%271511/1 · 0/02026-10-23gate-raised
≈alien-signals25.17 µs1.06×3.8%271411/1 · 0/02026-10-18stale: harness, stale: adapter, gate-raised
#5preact-signals194.50 µs8.16×1.5%271431/1 · 0/02026-10-18stale: harness, stale: adapter, gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

an application screen in use appRoundnarrow

The same app screen as "an application screen fills", mounted before timing starts. 20 rounds, each one step: every card's tag list grows by one tag, or drops from 4 back to one, mounting or unmounting tag components; every deck is renamed, every card toggled, and one user's setting changed. Mounts, unmounts and changes to cards, decks and users happen together. Each round must produce one update with exactly the expected components re-rendered.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array221.27 µs1.00×17.5%10515120/20 · 0/02026-10-31busy, gate-raised
≈oimdb-sorted-index341.73 µs1.56×2.6%10515520/20 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb351.62 µs1.59×2.4%10515520/20 · 0/02026-10-31busy, gate-raised
#4oimdb-3.12.1594.52 µs2.65×1.7%10511720/20 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, gate-raised
≈alien-signals647.86 µs2.82×3.9%10514120/20 · 0/02026-10-22busy, gate-raised
#6preact-signals857.48 µs3.74×9.7%10514320/20 · 0/02026-10-22busy, gate-raised
≈solid-js1.717 ms7.49×5.3%10514020/20 · 0/02026-10-22busy, gate-raised
≈vue-reactivity1.761 ms7.68×1.7%10517220/20 · 0/02026-10-22busy, gate-raised
#9mobx2.635 ms11×3.7%10515220/20 · 0/02026-10-22busy, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

one setting moves appSettingMoverepresentative

The same app screen as "an application screen fills", mounted before timing starts. 20 steps, each changing one user's setting. Only that user's 2 deck-date components read it and must update; nothing else may. Measured: how quickly a library finds the few components that depend on a value stored in another record.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array4.04 µs1.00×2.3%7,68015120/20 · 0/02026-10-22unresolved, busy, gate-raised
#2oimdb-sorted-index4.72 µs1.13×2.3%7,68015520/20 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈oimdb4.68 µs1.16×1.3%7,68015520/20 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
#4alien-signals5.66 µs1.39×4.6%7,68014120/20 · 0/02026-10-21unresolved, gate-raised
#5preact-signals6.73 µs1.65×3.3%7,68014320/20 · 0/02026-10-21unresolved, gate-raised
#6vue-reactivity8.95 µs2.20×1.0%7,68017220/20 · 0/02026-10-21unresolved, gate-raised
#7solid-js10.53 µs2.59×2.4%7,68014020/20 · 0/02026-10-21unresolved, gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a churn binding derives NO VIEW — nothing is subscribed until a script says so, and there is no window and no group in it anywhere. The group index this participant exists to price therefore does not exist in that wiring, and a churn row here would be `baseline-array`'s code registered under a second name: the same Map, the same subscriber arrays, the same indexOf and splice. Two identical rows read as two answers, so there is one.

read a page by key readByKeynarrow

8 steps. Each looks up 32 rows by key, spread across the whole store, and renders `id rank label` for each. Nothing is written: the only scenario that times reading rather than reacting to a change. The keys may not be looked up in advance. The hand-written baseline answers with one Map lookup per key.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array35.42 µs1.00×1.0%7,680898/8 · 0/02026-10-18stale: harness, gate-raised
≈baseline-indexed35.53 µs1.00×1.3%7,6801618/8 · 0/02026-10-22busy, gate-raised
≈oimdb-3.12.136.11 µs1.02×1.8%7,6801688/8 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, gate-raised
≈oimdb-sorted-index36.20 µs1.02×3.3%7,6801938/8 · 0/02026-10-22stale: adapter, stale: subject, unresolved, busy, gate-raised
≈oimdb36.31 µs1.02×2.4%7,6801938/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
#6preact-signals42.45 µs1.20×0.7%7,6802848/8 · 0/02026-10-18stale: harness, stale: adapter, gate-raised
#7alien-signals47.78 µs1.35×2.1%7,6803008/8 · 0/02026-10-18stale: harness, stale: adapter, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

scroll the window scrollWindownarrow

8 steps. Each scrolls view A within group A, a quarter page at a time down to one full page and back to the top. No data changes. Every step must show a new page; view B, a filtered view of the same group that does not scroll, must not change. Compare with "the whole window re-sorted": there the data moves under a fixed page, here the page moves over fixed data.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-3.12.183.12 µs0.092×2.1%1,8001628/8 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, busy, gate-raised
#2preact-signals42.36 µs0.43×0.8%1,8002768/8 · 0/02026-10-21gate-raised
≈alien-signals44.18 µs0.45×1.2%1,8002938/8 · 0/02026-10-21gate-raised
≈oimdb51.30 µs0.47×79.3%1,8001898/8 · 0/02026-10-31stale: harness, stale: adapter, stale: subject, busy, gate-raised
≈vue-reactivity51.19 µs0.52×4.1%1,8003018/8 · 0/02026-10-21gate-raised
≈oimdb-sorted-index53.00 µs0.53×1.3%1,8001898/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈solid-js65.34 µs0.66×14.8%1,8002848/8 · 0/02026-10-21gate-raised
#8baseline-array904.07 µs1.00×0.5%1,800838/8 · 0/82026-10-18stale: harness, busy, gate-raised
≈baseline-indexed610.34 µs1.00×177.2%1,8001558/8 · 0/02026-10-31stale: harness, busy, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

scroll deep into the group deepScrollnarrow

8 steps. Like "scroll the window", but view A jumps in four equal steps to the last full page of the group and back, as in virtual scrolling far from the top. No data changes. A library that keeps the group sorted jumps straight to the position; one that does not sorts the whole group again and discards everything above the page. Every step must show a full new page; view B must not change.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-3.12.189.16 µs0.087×1.3%1,2001628/8 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, busy, gate-raised
≈alien-signals48.67 µs0.23×2.6%1,2002938/8 · 0/02026-10-21gate-raised
#3preact-signals54.07 µs0.26×4.9%1,2002768/8 · 0/02026-10-21gate-raised
≈oimdb64.40 µs0.28×26.8%1,2001898/8 · 0/02026-10-31stale: harness, stale: adapter, stale: subject, busy, gate-raised
≈vue-reactivity61.01 µs0.29×1.9%1,2003018/8 · 0/02026-10-21gate-raised
≈solid-js76.28 µs0.36×3.5%1,2002848/8 · 0/02026-10-21gate-raised
#7oimdb-sorted-index111.14 µs0.53×10.3%1,2001898/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
#8baseline-array1.025 ms1.00×1.2%1,200838/8 · 0/82026-10-18stale: harness, gate-raised
≈baseline-indexed233.93 µs1.00×16.6%1,2001558/8 · 0/02026-10-31stale: harness, busy, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

the view switches group viewSwitchrepresentative

8 steps. Each switches view A to another group, like a user changing tab or filter: 7 different groups, then back to group A. No data changes. Every step must show a full page of the new group from its top; view B must not change. A library that keeps a list per group reads another list; one that does not filters the whole store again.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-3.12.174.56 µs0.079×5.9%8251648/8 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, busy, gate-raised
≈alien-signals51.30 µs0.61×8.8%8252968/8 · 0/02026-10-21gate-raised
≈preact-signals53.44 µs0.64×12.2%8252798/8 · 0/02026-10-21gate-raised
#4vue-reactivity61.30 µs0.73×3.8%8253028/8 · 0/02026-10-21gate-raised
#5solid-js80.41 µs0.96×2.7%8252878/8 · 0/02026-10-21gate-raised
≈baseline-array951.59 µs1.00×3.1%825848/8 · 0/82026-10-18stale: harness, gate-raised
≈baseline-indexed83.61 µs1.00×4.0%8251568/8 · 0/02026-10-22busy, gate-raised
≈oimdb-sorted-index116.74 µs1.40×5.5%8251948/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb119.29 µs1.43×5.1%8251948/8 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

a view filtered on two axes filterTwoAxesnarrow

4 steps with two views of the same group: view A filters by `group` only, view B by `group` and a second condition. The steps move two rows out of the group and back: one that matches view B's second condition and one that does not. View A must update all four times, view B only twice. Measured: how often view B was woken for a row its filter already excluded.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1preact-signals34.15 µs0.30×1.5%9002734/4 · 2/22026-10-21gate-raised
#2alien-signals37.28 µs0.32×4.2%9002904/4 · 2/22026-10-21gate-raised
≈vue-reactivity38.96 µs0.34×9.7%9002984/4 · 2/22026-10-21gate-raised
≈oimdb-sorted-index38.74 µs0.34×2.8%9001834/4 · 2/22026-10-22stale: adapter, stale: subject, busy, gate-raised
#5oimdb-3.12.1174.94 µs0.40×4.5%9001574/4 · 2/22026-10-18stale: harness, stale: adapter, stale: subject, busy, gate-raised
#6solid-js56.65 µs0.49×9.2%9002814/4 · 2/22026-10-21gate-raised
#7oimdb76.50 µs0.67×4.2%9001834/4 · 2/22026-10-22stale: adapter, stale: subject, gate-raised
#8baseline-array440.42 µs1.00×0.9%900794/4 · 2/42026-10-18stale: harness, gate-raised
≈baseline-indexed113.46 µs1.00×1.5%9001514/4 · 2/42026-10-22gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

relation lower bound relationFloorsynthetic

Lower bound, no library: the 16 steps of "a write to a related entity" and "a record keyed by two identities", each rendering the changed view straight from arrays of strings. Looking up the linked records is left out, because that is what those two scenarios measure. No library can come out below it.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array15.53 µs1.4%23,0401238/8 · 8/82026-10-18stale: harness, busy, gate-raised

a write to a related entity relationWritenarrow

8 steps with two record types: 8 rows each point to an owner, 4 owners with two rows each. Each step changes one owner's `name`; no row changes. View A shows each row with its owner's name, so it must update and two of its lines change. View B shows the same rows through another linked record and must not change. No row changes its owner.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb-sorted-index7.78 µs0.51×0.3%7,6802148/8 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈oimdb7.95 µs0.52×4.2%7,6802148/8 · 0/02026-10-31fine-unresolved, busy, gate-raised
#3mobx11.51 µs0.76×1.5%7,6802108/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#4preact-signals11.92 µs0.78×2.2%7,6802098/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈alien-signals12.03 µs0.79×13.4%7,6801918/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
≈vue-reactivity12.52 µs0.82×0.5%7,6802138/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#7solid-js14.33 µs0.94×0.8%7,6801768/8 · 0/02026-10-22fine-unresolved, busy, gate-raised
#8baseline-array15.29 µs1.00×1.9%7,6801548/8 · 0/82026-10-31fine-unresolved, busy, gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a relation binding derives NO VIEW — there is no window, no group and no sort in it anywhere, because what those two rows measure is a JOIN and a COMPOSITE KEY and a window derivation in front of either would be a constant sitting on both. The group index this participant exists to price therefore does not exist in that wiring, and a relation row here would be `baseline-array`'s code registered under a second name: the same maps, the same joined-string pair key, the same dirty flag. Two identical rows read as two answers, so there is one.

a record keyed by two identities compositeKeyWritenarrow

8 steps. Each changes the `mark` of a record identified by a pair (owner, row), with the two-part key built fresh on every call. View B shows these marks and must update every step; view A must not change. Measured: finding a record by a two-part key; a hand-written store joins the two ids into a string and looks it up in a Map. The records exist before timing starts, so nothing is created.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1oimdb9.11 µs0.54×4.0%7,6802140/0 · 8/82026-10-31fine-unresolved, busy, gate-raised
≈oimdb-sorted-index8.94 µs0.54×13.8%7,6802140/0 · 8/82026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
#3mobx14.21 µs0.87×2.6%7,6802100/0 · 8/82026-10-22busy, gate-raised
#4alien-signals14.61 µs0.89×1.8%7,6801910/0 · 8/82026-10-22busy, gate-raised
≈preact-signals14.79 µs0.90×0.3%7,6802090/0 · 8/82026-10-22busy, gate-raised
#6vue-reactivity15.80 µs0.96×1.1%7,6802130/0 · 8/82026-10-22busy, gate-raised
≈baseline-array16.82 µs1.00×3.0%7,6801540/8 · 8/82026-10-31fine-unresolved, busy, gate-raised
#8solid-js18.36 µs1.12×2.9%7,6801760/0 · 8/82026-10-22busy, gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a relation binding derives NO VIEW — there is no window, no group and no sort in it anywhere, because what those two rows measure is a JOIN and a COMPOSITE KEY and a window derivation in front of either would be a constant sitting on both. The group index this participant exists to price therefore does not exist in that wiring, and a relation row here would be `baseline-array`'s code registered under a second name: the same maps, the same joined-string pair key, the same dirty flag. Two identical rows read as two answers, so there is one.

single-value lower bound singleFloorsynthetic

Lower bound, no library: 32 renders of one line, each with a different value, with no store. Every single-value scenario renders through the same code, so the gap between a library and this bound is the library's own cost.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
≈baseline-array4.20 µs0.3%23,04012332/32 · 0/02026-10-21stale: harness, gate-raised

write one value singleWriterepresentative

32 steps. Each sets a new value on one standalone value: no key, no collection, no list. View A shows it and must update every step. A second value beside it never changes and must not update; how often its view was woken anyway is counted.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1baseline-array5.00 µs1.00×0.7%7,6804332/32 · 0/02026-10-22gate-raised
#2alien-signals5.87 µs1.17×2.7%7,6807632/32 · 0/02026-10-21stale: harness, stale: adapter, gate-raised
≈oimdb6.19 µs1.24×0.8%7,6809232/32 · 0/02026-10-22stale: adapter, stale: subject, gate-raised
≈oimdb-sorted-index6.22 µs1.24×3.3%7,6809232/32 · 0/02026-10-22stale: adapter, stale: subject, fine-unresolved, busy, gate-raised
≈preact-signals6.25 µs1.24×2.3%7,6807032/32 · 0/02026-10-21stale: harness, stale: adapter, gate-raised
#6oimdb-3.12.110.45 µs2.10×4.1%7,6806632/32 · 0/02026-10-18stale: harness, stale: adapter, stale: subject, gate-raised

Not yet measured: tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a single-value binding holds ONE VALUE and a second beside it — no rows, no keys, no groups and no window anywhere in it. The group index this participant exists to price is not merely unused there, it is not expressible: there is nothing to group. A single-value row here would be `baseline-array`'s code registered under a second name — the same two variables, the same two callbacks, the same dirty flag — and the `indexed − lean` line the report prints under a table both anchors appear in would be a reading of process noise presented as the price of an index. Two identical rows read as two answers, so there is one.

re-write one value singleResendnarrow

32 steps. Each sets one standalone value to the value it already has. The benchmark code does not filter this out, so it shows whether the library compares before it notifies: if it does, nothing renders; if not, the same line renders every step.

placeparticipanttime× anchordrifttimed runscode linesrenders A / Bmeasuredflags
#1alien-signals80 ns0.048×6.2%7,680760/0 · 0/02026-10-21stale: harness, stale: adapter, gate-raised
#2preact-signals85 ns0.051×5.0%7,680700/0 · 0/02026-10-21stale: harness, stale: adapter, gate-raised
#3oimdb-sorted-index409 ns0.25×1.9%7,680920/0 · 0/02026-10-22stale: adapter, stale: subject, busy, gate-raised
≈oimdb411 ns0.25×10.3%7,680920/0 · 0/02026-10-22stale: adapter, stale: subject, gate-raised
#5baseline-array1.66 µs1.00×0.8%7,680430/32 · 0/02026-10-22gate-raised

Not yet measured: oimdb-3.12.1, tinybase, signaldb, tanstack-db, tanstack-db-eager, tanstack-db-indexed, lokijs, mobx, jotai, legend-state, effector, nanostores, nanostores-per-row, redux-toolkit, zustand, valtio, xstate-store, vue-reactivity, solid-js, tanstack-store.

Declined by baseline-indexed

baseline-indexed: a single-value binding holds ONE VALUE and a second beside it — no rows, no keys, no groups and no window anywhere in it. The group index this participant exists to price is not merely unused there, it is not expressible: there is nothing to group. A single-value row here would be `baseline-array`'s code registered under a second name — the same two variables, the same two callbacks, the same dirty flag — and the `indexed − lean` line the report prints under a table both anchors appear in would be a reading of process noise presented as the price of an index. Two identical rows read as two answers, so there is one.

What each store holds​

The other half of the trade, in bytes: heapUsed after the whole default population has been written and both views delivered, minus the heap before bind(), both after two forced collections. There is one process and one reading per participant, so no drift and no rank. A store that keeps indexes pays for them here; one that derives by rescanning holds little and pays in the tables above. Each worker loads only its own library, so the baseline is the harness and nothing else.

participantheldper rowmeasured
baseline-array484.6 KB50 B2026-10-22
baseline-indexed1.02 MB107 B2026-10-22
zustand1.29 MB135 B2026-10-22
xstate-store1.32 MB138 B2026-10-22
redux-toolkit1.46 MB154 B2026-10-22
nanostores1.92 MB201 B2026-10-22
lokijs1.98 MB207 B2026-10-22
effector2.38 MB250 B2026-10-22
tinybase3.41 MB357 B2026-10-22
oimdb-3.12.14.03 MB423 B2026-10-22
oimdb-sorted-index4.23 MB443 B2026-10-22 · stale
oimdb4.26 MB447 B2026-10-22 · stale
signaldb4.40 MB461 B2026-10-22
preact-signals4.63 MB486 B2026-10-22
alien-signals5.51 MB577 B2026-10-22
tanstack-store5.85 MB614 B2026-10-22
tanstack-db5.89 MB618 B2026-10-22
vue-reactivity7.06 MB740 B2026-10-22
jotai9.85 MB1032 B2026-10-22
tanstack-db-eager10.04 MB1053 B2026-10-22
tanstack-db-indexed10.09 MB1058 B2026-10-22
nanostores-per-row10.74 MB1126 B2026-10-22
solid-js11.35 MB1190 B2026-10-22
valtio16.08 MB1686 B2026-10-22
mobx16.93 MB1775 B2026-10-22
legend-state22.92 MB2403 B2026-10-22

What one mechanism holds​

Bytes retained per row or per derivation, measured by npm run retained in packages/bench: the same population built in separate processes without the mechanism, with it, and without it again, and the heap difference (v8.getHeapStatistics().used_heap_size, after forced collections) divided by the row count. A positive number is what the mechanism costs. Nothing here is a timing.

Measured 2026-10-22 on node v24.16.0, Apple M1, 8 cores, 100,000 rows per row of the table.

what is measuredperbytesin referencesengine
a settled, held derivation (OIMComputedService, fan-in 3, value a number) over 100,000 entitiesderivation333.06 B41.65current core
an array of 3 per-source readings beside each derivation: the per-slot clock, modelled, not shippedderivation79.95 B10.00current core
the entities themselves: two strings, one number, two shared strings per rowrow171.91 B21.50core-legacy
detectExternalMutations: true against false: a shallow snapshot per key, its map, and a cleared Set's capacityrow100.79 B12.61core-legacy

How to read it.

  • What the derivation row includes. The slot and its fields (the two global clock stamps among them), the array of its three source references, its entry in the service's Map, one reference in the array holding it, and its numeric value. It does not include the entities, which are the same on both sides of the subtraction.
  • What it does not cover. A derivation with a different fan-in, or one whose value is a list (createOIMBucketMemoComputedService), is a different figure: the source array grows with the fan-in, and a list is dominated by its own array.
  • The per-slot readings row is a model. It prices the arm of the clock fork that does not ship, so it is the cost the shipped global stamps avoid, about a quarter of a derivation (79.95 / 333.06).
  • The two legacy rows. They weigh OIMCollectionChangedFields in packages/core-legacy, not this engine. Read together, the snapshot option adds 100.79 B to rows that hold 171.91 B on this row shape. A shape with more bytes behind its references moves the entity row up and leaves the snapshot row where it is.

What the instrument was checked against, in the same run.

  • Bytes per reference. A pre-sized array of 1,048,576 references measured 8.00 B per reference, and one of 131,072 measured 7.96 B. That is the uncompressed answer: this build has no pointer compression.
  • "In references". That column is the per-row figure divided by 8.00 B.
  • Noise floor. The same 100,000-row population built twice, in two processes, came out 64 B apart. No difference smaller than 64 B is a reading.