One value per two bytes of the compressed canvas was too tight for a
public design system: a 39.5 MB canvas holding 29.7 million values got
a budget of 19.7 million (reported in #701). Real files hold 0.37 to 0.75
values per compressed byte; the budget is now one per byte, at least the
old 20 million and at most 128 million. A small file still gets main's
20 million, however far it unpacks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
`document.json` was read under the 512 MB cap for any archive member, so a
large imported document saved but didn't open again: 900 000 layers save
as a 1.48 GB document.json, packed into 244 MB. It may now unpack to 16
times its packed size (counted up to the archive's own size), within
512 MB and 4 GB; images and other members keep 512 MB. A small archive
still can't unpack into gigabytes.
That saved file reopens in 13 s with a 5.9 GB peak.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
A real 705 MB .fig file (401 644 nodes, 894 004 layers once instances are
expanded) didn't open: "message: too many values". Lifting that cap alone
made it import only its first 200 000 layers, and lifting that one took
41 GB of memory.
The global caps stay as they are. The .fig import gets budgets from the
size of its compressed canvas instead (counted up to the file's own size),
not from the message it unpacks to, since chunks inflate up to 512 MB:
- the kiwi value budget is one value per 2 bytes of the canvas, at least
the old 20 million (kiwi is only read for .fig files);
- the layer cap is one layer per 32 bytes of the canvas, at least the REST
import's 200 000 and at most 2 000 000. A small or crafted file can't
expand its instances into millions of layers, however far it unpacks; the
REST JSON import keeps 200 000.
Memory:
- The mapper no longer passes the document through `json!`, which serialises
its arguments: the whole tree was deep-copied twice.
- Vector networks are built once per distinct network and handed to the REST
importer beside the JSON (`vectorNetworkRef`), not written into it as JSON
objects for every instance that shows them.
- Decoded kiwi objects are trimmed, and each import stage's input is dropped
as soon as the next stage is built.
On that file: peak footprint 41 GB -> 11.3 GB, 67 s -> 25 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
#639 lands first and reads networks with decode_styled; once #631 drops the count it calls
network::decode. Needed whenever #631 is rebased onto a main that has #639.
Image filter warnings name fields of the file, so a crafted schema could add any number of
distinct warnings (each checked against all earlier ones). The mapper keeps at most as many as
the REST importer does.
The Typography section's variable picker also lists number variables for a text layer's content (shown as the number), as the engine already accepted; other text fields keep their own types (#629).
Both carry the same image filter value, which the public docs describe as green when lower and
magenta when higher; a positive native tint adds green. Both readers flip it now (the .fig path
did, the REST path didn't). The .fig test message no longer names the product.
The test schema extended the image scale-mode field that only #634 adds; it now adds the
filter containers to the paint message itself, so it works with and without #634.
Image paints that carry a current filter container now import their exposure, contrast, vibrance (as saturation), temperature, tint, highlights and shadows as native image filters. Neutral values leave the paint untouched. The older color-adjust container alone, both containers together, filters that have no native equivalent (brightness, detail, vignette and unknown fields) and malformed or non-finite values are skipped and reported as import notes. A positive tint in .fig data moves toward magenta, so it is flipped once at the .fig boundary; public REST image filters keep their sign and are not reported as approximated. Imported filters use the approximate native renderer, which the import says once.
Supported TEXT_PATH layers import as editable native text on their vector path, including mixed text styles, a start inside a later segment and reversed text direction. The importer converts the segment parameter stored in the file to the native distance along the complete path.
The mapping accepts a connected, unbranched path with one recognized start-segment marker and the default placement above the path. Damaged or unsupported path data keeps the text readable off its path and adds an import note. Other vertical placements, multiline text and ambiguous or disconnected networks are not converted. Underline and strikethrough remain a native path-rendering limitation and are reported. Public REST TEXT_PATH layers and the experimental .fig writer are unchanged.
The vector network decoder now keeps each segment's style id. Import reports a segment style that names a row of the layer's style table as metadata that was not interpreted, and one that names no row as a reference to a missing style entry, each at most once per file. The earlier note claimed that stroke styles were left out, which overstated what is known about the metadata. Segment geometry, paint and width are unchanged, and the style ids are not kept in the native vector model.
The public docs don't give a default for skip ink; they say the value is unset when the text
isn't underlined, and the exporter always wrote it on. So the export keeps writing on for text
without an underline and the real value under one (that's what the canvas draws), and the import
leaves a value on text without an underline out of the decoration look, so plain text keeps the
default look and round-trips unchanged. The native default stays off, so existing documents
draw as before.
Plugins can no longer reach the disk or network through path parameters (variable.export/codeConnect.export/component.playground writes, codeConnect.import/variable.import reads) or through collab.* and window.*; found and fixed by @lucasmartins-ai, split out of #696.
One value per two bytes of the compressed canvas was too tight for a
public design system: a 39.5 MB canvas holding 29.7 million values got
a budget of 19.7 million (reported in #701). Real files hold 0.37 to 0.75
values per compressed byte; the budget is now one per byte, at least the
old 20 million and at most 128 million. A small file still gets main's
20 million, however far it unpacks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
DocState::new already settles every opened document (native files, branches and live joins
included); the .fig/REST path keeps its settlement in link_instances for file.import into an
open document.
Frames on the page or in a section convert to sections (object.convertToSection) and sections back to frames (object.convertToFrame), in place, one undo step each, from the canvas context menu; nested frames, instances, main components and sections holding sections are refused.
bflatastic