12 Commits

Author SHA1 Message Date
Álvaro Tejero-Cantero
678fcc071b
🐛 Check the page that an add-page change stores (#12675)
* 🐛 Check pages added with a full page value in soft mode

An :add-page change can carry a whole page under :page instead of an
id and a name. extract-affected-ids reads the page to check from the
change's top-level :id, so that form left the new page unchecked and
a referential error in it was stored.

Collect those page ids after the changes are applied and run the
affected check on them in soft mode: log the file id, the page ids
and the error codes, and never refuse the save. This is the first
step of the two-step rollout. The follow-up moves the page id into
extract-affected-ids and makes the check hard. The counters of
#12665 decide when to flip it.

Relates to #12666

AI-assisted-by: claude-opus-5-5, deepseek-v4.1-flash
Signed-off-by: Álvaro Tejero Cantero <alvorithm@teje.ro>

* 🐛 Collect the page of a whole-page add-page in the affected check

An `:add-page` change that carries the page alone, with its id inside
`:page` and none at the top level, escaped the referential check:
`extract-affected-ids` read only the top-level `:id`. The schema side
(`touched-page-ids`) already reads `(or id (:id page))`; the referential
side now does the same.

This replaces the separate soft-mode check of the first commit. The page
now goes through `validate-file-affected` like any page a save touches,
so the deployment's flags decide whether an invalid page is refused
(`:file-validation`) or only logged (`:soft-file-validation`), and no
second log hint or rollout step is needed.

AI-assisted-by: claude-opus-5-5

* 🐛 Check the page an add-page stores, not its top-level id

An :add-page that carries the page whole stores it under the id inside
:page. extract-affected-ids and touched-page-ids read the top-level :id
first, so a change whose two ids differ stored one page and checked
another, which does not exist. Both now read the id of the page first.

The unit test of the affected check moves to the common validate tests,
with a case for the differing ids on the referential and the schema
side. The backend test sends that form through update-file as well,
and a new test sends the page as a commit group, the way the editor
does, and expects the commit under :rejected and no stored page.

AI-assisted-by: claude-fable-5-2

* ♻️ Derive the add-page change ID from the page

Use the page ID for both redo and undo so the builder cannot emit
conflicting IDs. Keep the change format and the library API unchanged.

AI-assisted-by: gpt-6.2-sol

---------

Signed-off-by: Álvaro Tejero Cantero <alvorithm@teje.ro>
Co-authored-by: Andrey Antukh <niwi@niwi.nz>
2026-11-20 17:47:53 +01:00
Andrey Antukh
b6a08ea4cb
⚡ Fuse the reg-objects of a change batch per container (#12310)
The component sync emits a `:reg-objects` for the parents of every
shape it touches, so a component swap collected 291 of them for 36
containers. Recomputing a group's geometry N times in a row gives the
same result as doing it once, so most of that work was thrown away,
in the browser and on the server, on every save.

`mch/merge-reg-objects` collapses them into one per container, with
the union of the shapes. The merged change moves to the end of the
batch: a group has to be recomputed after the shapes that resized
it, not before, or it keeps a stale `:selrect` until something else
touches it. That is also why it must not stay in the position of the
first change it replaces.

It runs on every commit, redo and undo alike, so the undo stack gets
the same benefit, and again when the buffer is batched for saving.
Measured on the swap that motivated this: 291 `:reg-objects` become
36, and the batch goes from 806 changes / 6.4 MB to 551 / 324 KB.

AI-assisted-by: space-bunny
2026-11-20 13:50:17 +01:00
Andrey Antukh
fd785b8246
⚡ Stop resending deleted component objects on every touch (#12308)
* ⚡ Stop resending deleted component objects on every touch

`pcb/update-component` attached the whole `:objects` map of a deleted
component to every `mod-component` change, so a single component swap
re-sent thousands of shapes that the same batch already carried as
shape changes. In one measured swap the 49 maps with objects were 94%
of the request (6.3 MB), and only 27% of the 4.275 shapes they carried
differed from what the server had.

Now `mod-component` carries component metadata only, and shapes inside
a component travel as `:mod-obj` changes scoped with `:component-id`,
which is what the component sync already emits. An absent `:objects`
key means "leave the stored shapes alone", only an explicit `nil`
drops them, so metadata-only changes cannot wipe a deleted component.

The two repairs that fix shapes inside a component (shape-ref cycle
and duplicated swap slot) now emit one `:mod-obj` per shape; the two
that move or clear the `:objects` key keep the whole-map write behind
the new `:objects? true` option.

Same swap on the measured file: 6.328.307 to 382.571 bytes.

AI-assisted-by: space-bunny

* 🔧 Cover the mod-component objects contract with tests

`pcb/update-component` sends the `:objects` map only when the caller
passes `:objects? true`, and `ctkl/mod-component` reads an absent key as
"leave the stored shapes alone". Two repairs depend on that option, and
dropping it broke nothing visible: the change carried no `:objects`, the
applier kept the invalid component as is, and the suite stayed green.

Add one repair test per error code, including the deleted-component
branch that must keep the map empty, and a round-trip test for the
shape-ref cycle repair. That repair now emits one `:mod-obj` per shape
and had no coverage; the test pins the scoped `:component-id` and that no
shape is lost.

Reword the `update-component` docstring as well: it describes what the
option does to the change instead of listing its current callers, which
goes stale as soon as another caller uses it.

AI-assisted-by: deepseek-v4.1-flash

* 🔧 Harden the mod-component repair tests

Address the review of the previous commit.

The cycle test only rejected a shape-ref pointing to itself, so it would
have passed if the repair repointed the link instead of removing it, or
if it left the component links on the main root untouched. It now
compares the resulting objects map with the expected one.

The four tests only inspected the first change, so a duplicated change
went unnoticed. They now compare the full list of changes: an exact
vector for the two repairs that rewrite the map, and id frequencies for
the cycle repair, which emits one change per shape in no fixed order.

The deleted-component fixture now removes the main instance through
`tho/delete-shape`, the same change generator the editor uses, instead
of a hand-written `:del-obj`. The message that promised a restorable map
now says what the repair does: normalize the value to an empty map.

AI-assisted-by: deepseek-v4.1-flash
2026-11-18 23:18:28 +01:00
Belén Albeza
fa1b3d4708
⚡ Keep only the shape data the worker reads (#12840)
* ⚡ Keep only the shape data the worker reads

The worker kept a full copy of the current page for its selection and
snap indexes. It now keeps only the attributes the indexes and the
change replay read, and re-indexes only the shapes that changed.

On a 52k-shape page the worker heap goes from 698 MB to 286 MB.

Closes #12740

AI-assisted-by: claude-opus-5-5

* ♻️ Share the worker re-trim with its replay test

The worker and its replay test now share `retrim-objects`.

AI-assisted-by: claude-opus-5-5
2026-11-05 16:25:29 +01:00
Eva Marco
7e3f779d8c
🎉 Add tokens to the canvas (#11923)
* ✨ Show color tokens toggle on canvas background picker

* 🐛 Apply token click on canvas background with no shape selected

* ✨ Apply and persist color tokens on canvas background

* 🌐 Translate canvas background section label
2026-09-28 08:53:42 +02:00
Alonso Torres
85bba64209
🐛 Fix and check idempotency in changes (#11823) 2026-09-23 16:09:11 +02:00
Andrey Antukh
81c3b3cd56
📎 Update copyright name on file header (#11346) 2026-08-25 11:55:10 +02:00
Andrey Antukh
7eb188b077
🐛 Fix text content repair and remove blast-radius containment (#10731)
Repair text shapes with empty/broken content at all three levels
(root, paragraph-set, paragraph) in migration 0025 to prevent
workspace update failures. Handle all corner cases: nil/empty/non-
vector children, non-map items, wrong types. Remove geometry-only
validation skip in changes.cljc so all shapes are validated.

AI-assisted-by: qwen3.7-plus
2026-07-28 11:00:53 +02:00
Alonso Torres
ca81776d04
🐛 Fix problems when dragging frame with comments (#10460)
* 🐛 Fix undo frame position not undoing comments

* 🐛 Fix problem with hover capturing dragging event

* 🐛 Fix watch updates for comment bubbles

* 🐛 Fix "Maximum update depth" crash on SVG shape transforms

* 🐛 Fix comment geometry problems
2026-06-30 13:33:47 +02:00
Yamila Moreno
ddba2ffa75
📎 Update Kaleidos Copyright (#9929) 2026-05-29 11:24:58 +02:00
Andrés Moya
d76a5c615c 🔧 Modify token sets by id instead of name and review usage 2025-09-29 12:16:42 +02:00
Andrey Antukh
21dd9a260c 📎 Rename common files changes test namespace 2024-09-18 12:15:12 +02:00