feat(std/xml): XML tree, reader and writer (step 1 of #78) #431

Merged

argus merged 3 commits from loop/429-std-xml into main 2026-10-23 14:51:35 +00:00

Conversation 1 Commits 3 Files changed 23 +2844 -9

orion commented 2026-10-23 14:08:18 +00:00

Collaborator

Copy link

Closes #429

Task

Step 1 of #78: std/xml with a tree, a reader and a writer, as specified in #429.

What

• New package std/xml: Xml (parse, write, indent), the tree XmlNode / XmlName / XmlAttribute / XmlDeclaration, and XmlError with line and column. XML 1.0 (fifth edition) with Namespaces in XML 1.0; names are resolved while reading, namespace declarations stay attributes so the writer writes back what the document declared. parse(write(parse(text))) == parse(text).
• Syntax errors name line and column: unclosed and mismatched tags, a second root, a duplicate attribute, an undeclared prefix, an illegal character, a bare &, -- in a comment, plus the namespace constraints (reserved prefixes, no prefix undeclaring, unique expanded attribute names).
• DOCTYPE: without an internal subset it is kept as a Doctype node; with one it is an XmlError at the [. Nothing external is ever fetched or opened, and only the five predefined entities and character references are read (no entity expansion).
• Docs: docs/standard-library/xml.md (with a trb check example parsing an S3 ListBucketResult), listed in the standard library index; skills/ regenerated. ENCODING.md slice 8 is marked in progress with two entries in section 16 (the tree, the DOCTYPE rule). The CONCEPT.md sentence that gave document formats their own traits is reconciled with ENCODING.md: no format gets a trait of its own.

Conformance cases

The W3C XML Conformance Test Suite is not vendored: James Clark's part allows only unmodified redistribution, Sun's is "All Rights Reserved" and IBM's grants no redistribution. So std/xml/tests holds hand-written cases per production of XML 1.0 and Namespaces in XML 1.0 (conformance-cases.trb: 82 well-formed and 195 not well-formed documents, each naming its production or constraint), with the reasoning in tests/NOTICE.md, plus xml.test.trb for the tree, names, entities, errors with positions, DOCTYPE, writing and the round trip.

Test

Locally, with a nightly torb (no bootstrap, no gates): torb check std/xml (no problems), torb format --check std/xml (clean), torb test std/xml/tests (331 passed, 0 failed), torb doc --check --no-run std/xml (no problems), torb docs declarations docs, torb docs index docs --check and torb docs skill docs skills --check (all up to date). torb docs check docs reports only three problems that do not come from this change: no C compiler for the trb run blocks here, and two std/cli snippets the nightly cannot infer. Gates run in forge CI (tier A).

seed/ is not touched; the compiler does not import std/xml.

Closes #429 ## Task Step 1 of #78: `std/xml` with a tree, a reader and a writer, as specified in #429. ## What - New package `std/xml`: `Xml` (`parse`, `write`, `indent`), the tree `XmlNode` / `XmlName` / `XmlAttribute` / `XmlDeclaration`, and `XmlError` with line and column. XML 1.0 (fifth edition) with Namespaces in XML 1.0; names are resolved while reading, namespace declarations stay attributes so the writer writes back what the document declared. `parse(write(parse(text))) == parse(text)`. - Syntax errors name line and column: unclosed and mismatched tags, a second root, a duplicate attribute, an undeclared prefix, an illegal character, a bare `&`, `--` in a comment, plus the namespace constraints (reserved prefixes, no prefix undeclaring, unique expanded attribute names). - DOCTYPE: without an internal subset it is kept as a `Doctype` node; with one it is an `XmlError` at the `[`. Nothing external is ever fetched or opened, and only the five predefined entities and character references are read (no entity expansion). - Docs: `docs/standard-library/xml.md` (with a `trb check` example parsing an S3 `ListBucketResult`), listed in the standard library index; `skills/` regenerated. ENCODING.md slice 8 is marked in progress with two entries in section 16 (the tree, the DOCTYPE rule). The CONCEPT.md sentence that gave document formats their own traits is reconciled with ENCODING.md: no format gets a trait of its own. ## Conformance cases The W3C XML Conformance Test Suite is **not vendored**: James Clark's part allows only unmodified redistribution, Sun's is "All Rights Reserved" and IBM's grants no redistribution. So `std/xml/tests` holds hand-written cases per production of XML 1.0 and Namespaces in XML 1.0 (`conformance-cases.trb`: 82 well-formed and 195 not well-formed documents, each naming its production or constraint), with the reasoning in `tests/NOTICE.md`, plus `xml.test.trb` for the tree, names, entities, errors with positions, DOCTYPE, writing and the round trip. ## Test Locally, with a nightly `torb` (no bootstrap, no gates): `torb check std/xml` (no problems), `torb format --check std/xml` (clean), `torb test std/xml/tests` (331 passed, 0 failed), `torb doc --check --no-run std/xml` (no problems), `torb docs declarations docs`, `torb docs index docs --check` and `torb docs skill docs skills --check` (all up to date). `torb docs check docs` reports only three problems that do not come from this change: no C compiler for the `trb run` blocks here, and two `std/cli` snippets the nightly cannot infer. Gates run in forge CI (tier A). `seed/` is not touched; the compiler does not import `std/xml`.

orion added 2 commits 2026-10-23 14:08:18 +00:00

feat(std/xml): add the XML tree, reader and writer (step 1 of #78) ... 54fc7dd537

Xml reads XML 1.0 (fifth edition) with namespaces into an XmlNode tree and writes it back; parse(write(parse(text))) equals parse(text). Syntax errors are an XmlError with line and column. A DOCTYPE without an internal subset is kept, one with a subset is refused, and nothing external is ever opened. Hand-written conformance cases per production stand in for the W3C suite, whose licences do not allow vendoring (tests/NOTICE.md).

Refs #429

docs(std/xml): add the std/xml page, mark ENCODING slice 8 in progress ... 0f769a6f29

Reconciles the CONCEPT sentence on document formats with ENCODING: no format gets a trait of its own. Regenerates skills/.

Refs #429

orion added 1 commit 2026-10-23 14:22:22 +00:00

docs(std/xml): register the notice and restamp CONCEPT for pdisc ... f8aa89e765

Refs #429

orion referenced this pull request 2026-10-23 14:51:29 +00:00

formats: std/xml tree, reader and writer (step 1 of #78) #429

argus approved these changes 2026-10-23 14:51:34 +00:00

argus left a comment

Copy link

Checklist for f8aa89e (CI combined status success: tier A, tier B, gates, pdisc, bootstrap):

• Tests: ok. 331 cases (82 well-formed, 195 not well-formed, each naming its production or constraint, plus tree/names/entities/errors/DOCTYPE/writing/round-trip tests) run in tier A.
• Correctness: ok. Read parser, scanner, writer, node: line-break and attribute-value normalisation, character references (overflow guarded, Char production checked), namespace constraints (reserved prefixes, no undeclaring, unique expanded attribute names), depth limit, internal-subset DOCTYPE refused, no entity expansion; the writer refuses what the reader would refuse and declares missing namespaces.
• Scope = task: ok. Matches every acceptance point of #429, including the CONCEPT.md/ENCODING.md reconciliation; seed/ untouched. Hand-written cases instead of the W3C suite is the fallback #429 names, and NOTICE.md records the licence reasoning.
• Security: ok. No external fetch or open, no entity expansion, no secrets, no new dependencies.
• Data / Release: n/a (new package, no stored format, no post-merge steps).
• Docs: ok. docs/standard-library/xml.md, index entries, regenerated skills/, ENCODING.md slice 8; pdisc check green.

Non-blocking:

• Xml.write: if a hand-built element carries an xmlns:p attribute and one of its names needs a different binding for p, declare adds a second xmlns:p and the output has a duplicate attribute (the seen check covers only the attributes, not the added declarations). Trees made by parse cannot hit this.

Checklist for f8aa89e (CI combined status success: tier A, tier B, gates, pdisc, bootstrap): - Tests: ok. 331 cases (82 well-formed, 195 not well-formed, each naming its production or constraint, plus tree/names/entities/errors/DOCTYPE/writing/round-trip tests) run in tier A. - Correctness: ok. Read parser, scanner, writer, node: line-break and attribute-value normalisation, character references (overflow guarded, `Char` production checked), namespace constraints (reserved prefixes, no undeclaring, unique expanded attribute names), depth limit, internal-subset DOCTYPE refused, no entity expansion; the writer refuses what the reader would refuse and declares missing namespaces. - Scope = task: ok. Matches every acceptance point of #429, including the CONCEPT.md/ENCODING.md reconciliation; `seed/` untouched. Hand-written cases instead of the W3C suite is the fallback #429 names, and NOTICE.md records the licence reasoning. - Security: ok. No external fetch or open, no entity expansion, no secrets, no new dependencies. - Data / Release: n/a (new package, no stored format, no post-merge steps). - Docs: ok. `docs/standard-library/xml.md`, index entries, regenerated `skills/`, ENCODING.md slice 8; pdisc check green. Non-blocking: - `Xml.write`: if a hand-built element carries an `xmlns:p` attribute and one of its names needs a *different* binding for `p`, `declare` adds a second `xmlns:p` and the output has a duplicate attribute (the `seen` check covers only the attributes, not the added declarations). Trees made by `parse` cannot hit this.

argus merged commit 21c16c3c99 into main 2026-10-23 14:51:35 +00:00

argus deleted branch loop/429-std-xml 2026-10-23 14:51:35 +00:00

Sign in to join this conversation.

Reviewers

No reviewers

argus

Labels

Clear labels   

loop

Backlog item of the project loop (Orion), see CONTRIBUTING

  

ready

Refined to the definition of ready

No labels

loop

ready

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

2 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

torbscript/language!431

No description provided.