Repository navigation
Spec 001 T035: compare - #173
Merged
Merged
Conversation
pdf-export compare <reference> <new> [--expected <file>], as contracts/cli.md has it: each side a PDF or a folder, a folder compared file by file by name. Both sides are read with pdf.js (lines, links, figures as clusters of images and painted paths) and reduced to FR-012's parts: the cover's title, diagram, excerpt note and distinct names (counted and hashed; U+2060 and no-break spaces normalised); the introduction's release and counts; the analysis's properties; each section's parts by their labels and their text; the contents' entries and targets; each figure's kind. The literature and edit history, which name people, are held as hashes. The JSON report gives each difference by part, field and kind with a count and stable identifiers, list order as a kind of its own; no section text and no name is printed. --expected takes the recorded differences, with their reasons. Tests print a sample document with pdf-export's layout and as a reference in event-pdf's shape (placeholder names, wording written anew): each part's difference is caught, and no name appears in any report. 21 mutants, each caught. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t strict shapes One printer (src/compare/output.ts) every reported value passes through: text read from a PDF is printed only as a count, yes or no, a release, one of a fixed list, a stable identifier, an analysis token of the service's form or one of Reactome's species; anything else as a hash and a length, salted once a run. Tested with names planted everywhere in 300 random layouts a side. - H1: the cover title is a hash, said only to be its file's name or not. - H2: the analysis part ends at the first stable identifier or heading, and a property is kept only of its shape. - M1: a contents target not read on either side is missing-target. - M2: links, figures and targets read by page; 4,000 sections in 1.25 s (30.9 s before on the same layout); a file's bytes not copied twice. - M3: the sections' order by their longest common subsequence, each occurrence of a repeated section compared with its own. - Lows: --reference-margin; an unreadable file named and a difference; --expected refuses unknown keys, keeps an events record to differences with events, and lists records matching nothing; docs say what is done. A second reference layout, unlike the sample's, found that a thin connector must join a vector figure's shapes. 38 mutants, all caught. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- M1: pdf.js's warnings quoted a PDF's content on standard error. Every getDocument reading a PDF that may hold names (compare, check-book, the contents' reader) now runs with verbosity ERRORS, and compare holds the console while it compares: a library's messages are counted and said once with their errors' kinds, never their words. Tested with a PDF whose content holds a planted word as an unknown operator. - M2: a table's hairline borders joined its shaded header into a figure and its rows were lost. A line of thin paths now joins clusters only where it touches two or more node-sized ones, as a connector does. - Lows: assertions that kill #167's surviving mutants; a missing path is exit 2 in the contract; the token is printed in its canonical form; the common subsequence's quadratic case recorded. 10 mutants, all caught. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…by kind - M1: a connector clipped at a reaction's small square was segments touching neither each other nor two nodes, so the diagram was no figure and its labels were read as text. Shapes under 4 pt are now pieces of lines like thin paths, so the segments and the square are one line, which joins the nodes at its ends; a shaded-header table still reads its rows. - Found with it: the last section's figures were sought on its first page only, and the first section's heading could reach onto the cover. - L1: pdf.js's errors quote the file. check-book's and the contents' readers turn them into UnreadPdf (src/print/unread.ts), said by kind alone, and check-book's stopped line prints kind and code only. - L2/L3: every console method is held during a comparison, put back on failure too, the calls said in one line; each tested. - L4/L5: the wording of what joins, and of tables still read as figures. 11 mutants, all caught. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…st can fail - M1: check-book's stopped line printed the wrapper's kind alone, and fs and graph errors lost their path or message. UnreadPdf now names its file and pdf.js's error's kind (kept as `kind`, pdf.js's error as its `cause`, never printed); the stopped line prints the message, so fs and graph errors say theirs. Tested with a planted filter, a truncated PDF and a file it may not read. - M2: the contents reader's test now reads a PDF whose destinations pdf.js throws on quoting the planted word, so a message carrying the cause's fails it. - Lows: a figure three pages into the last section; only the console's writing methods held; landed's doc comment restored; a shaded header just below a figure among the known gaps. 7 mutants, all caught. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.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.
Spec 001, task T035. Replaces #171 after its review (two mediums, fixed in the second-last commit); #171 replaced #168, #167 and #165.
Everything in #171 (
pdf-export compare; nothing read from a PDF printed unless of a known shape; pdf.js quiet;UnreadPdffor PDF read failures), plus, from #171's review:UnreadPdf, no file): it now names the file and pdf.js's error kind (<path> not read (InvalidPDFException)), keeping pdf.js's own error as an unprintedcause; file-system and graph errors print their own message, which holds no PDF text.UnreadPdf's message doesn't.Tests:
test/compare/(59),test/check/unread.test.ts(4); 518 in all. Mutants in tasks.md.🤖 Generated with Claude Code