Repository navigation
build(ci): update dtolnay/rust-toolchain requirement to 7e38f4b43b4db5c8dd498af069a4f6196df1d067 - #46
Open
dependabot[bot] wants to merge 1 commit into
Conversation
…5c8dd498af069a4f6196df1d067 Updates the requirements on [dtolnay/rust-toolchain](https://github.com/dtolnay/rust-toolchain) to permit the latest version. - [Release notes](https://github.com/dtolnay/rust-toolchain/releases) - [Commits](https://github.com/dtolnay/rust-toolchain/commits/7e38f4b43b4db5c8dd498af069a4f6196df1d067) --- updated-dependencies: - dependency-name: dtolnay/rust-toolchain dependency-version: 7e38f4b43b4db5c8dd498af069a4f6196df1d067 dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com>
Author
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
ethos-pramaan
pushed a commit
that referenced
this pull request
Oct 6, 2026
…drawing, decision #46 docs/34-FIGURE-REGIONS-SCOPE.md. 819 of ParseBench's 2,202 pictures had no element of ours overlapping them at all: charts, diagrams and logos drawn with paths rather than placed as images. The owner, shown a prototype's measurement, chose "Write scope, then build". The rule: every path a fill or a stroke paints, on the page and in the forms it draws, less any whose box covers 80% of the page, clustered wherever the boxes come within 3 points across and down. A cluster is a region where it holds 3 paths or more, covers 0.1% to 80% of the page, has less than half its area inside any one table, and holds no baseline of more than 60 characters. Only on a document that declares no author structure. A region claims no run and neither projection writes it. On the wire: RepresentationPayload gains a tenth member, `figures`, always written, so schema_version 0.7.0 -> 0.8.0 and both packages read 0.8.0. Capabilities gains `figures` (true for PDF; the office profiles declare figures-not-detected), the profile gains `figure_rule`, and profile_sha256 moves 9ed655a7... -> d6531270.... Each FigureRecord is its page, its box within the page, Computed, and the rule id; no id of its own, as an outline record has none. New engine fixture `figure-paths-drawn`: a chart drawn with axes, bars and a curve inside a clip that `n` ends unpainted, beside a lone rule. 18 mutations over the clustering, the four keep-clauses, the page cut, the order and the interpreter's path boxes are all killed. Measured on ParseBench: visual grounding 0.4708 -> 0.4824, 73 pages up and 3 down by at most 0.036 (running heads drawn as tab bands); content, formatting, tables and charts unchanged on every page; overall 42.55 -> 42.78. opendataloader-bench: all 200 Markdown files byte-identical. The gate documents are tagged, so nothing runs on them; their tree-stripped copies carry 208 regions over 1,466 pages. Signed-off-by: docushell-dev <hello@docushell.com>
ethos-pramaan
pushed a commit
that referenced
this pull request
Oct 6, 2026
…s unit After decision #46, 1,274 of the Text elements ParseBench's visual grounding failed still had a predicted item more than half as tall again as they were: a contents list, a column of labels, a list set without markers, whose lines end short of their unit's widest by a word or two, too little for line-units-v2's ten line heights. The rule: after -v2's cuts, a unit of body text is also cut before a line whose first word, with the space before it, would have fit twice over between where the line above ends and the unit's widest piece's right edge. The measure did not break that line, so its author did. The first word's width is read at its own piece's width per character. A heading unit is not cut, so the projections still join a heading set on two lines. Measured first as a split of the adapter's items: the first word fitting once / twice / three / four / six times over scored 0.4748 / 0.4822 / 0.4798 / 0.4780 / 0.4743 against 0.4708. In the engine, cutting every unit parted two-line headings (formatting 0.4080 -> 0.4072); cutting body text only is what ships. The same test the other way -- a wrapped line joining across up to one line height -- was measured and refused (0.4817 -> 0.4771). layout_unit_rule line-units-v2 -> line-units-v3, so profile_sha256 moves d6531270... -> d5e4f162.... New engine fixture `units-contents-list`: a two-line heading the cut must leave whole, a wrapped paragraph, and a contents list -v2 read as one unit, cut into its entries. 11 mutations over the clause, the first-word estimate and the wiring are all killed. Measured on ParseBench with decision #46: visual grounding 0.4824 -> 0.4933, 72 pages up and 49 down (332 Text elements pass that did not, 87 no longer do, mostly a list the ground truth keeps as one element); content, formatting, tables and charts unchanged on every page; overall 42.78 -> 43.00. opendataloader-bench: all 200 Markdown files byte-identical. Signed-off-by: docushell-dev <hello@docushell.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Updates the requirements on dtolnay/rust-toolchain to permit the latest version.
Commits
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)