Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[0.0.90] - 2026-11-16

Changed

  • Release 0.0.90 in lockstep with the pain001 suite.

  • The container image moves from python:3.12-slim to python:3.15-slim (Dependabot #115), pinned by digest in both Dockerfile stages. The test matrix stays on Python 3.11 to 3.14.

  • Requires pain001 0.0.90 (floor >=0.0.90,<1), the suite version. pain001 0.0.90 keeps 0.0.89’s dependency floors, so no other package moves in the lock or the hash-pinned sets.

[0.0.89] - 2026-11-16

Changed

  • Release 0.0.89 in lockstep with the pain001 suite.

  • Requires pain001 0.0.89 (floor >=0.0.89,<1), the suite version. Its own floor on click 8.5.1 moves click to 8.5.1 in the lock and the hash-pinned sets.

  • Dependency updates folded in from Dependabot (#110 to #114): ruff 0.17.5 (floor >=0.17.5 in pyproject.toml and requirements/ci.in), complexipy 9 (range >=9.0.0,<10), filelock 4.2.10 in the ci, audit and build sets, and myst-parser 5.2.0 with mdit-py-plugins 0.7.0 in the docs set.

  • complexipy 9 scores cognitive complexity higher than 8 did and put scripts/check_complexity.py’s compare at 18, over the ceiling of 15. It is split into _regressions and _improvements with the same output and order; the complexity baseline stays empty.

[0.0.88] - 2026-11-14

Changed

  • Release 0.0.88 in lockstep with the pain001 suite.

  • Requires pain001 0.0.88 (floor >=0.0.88,<1), the suite version.

  • Suite Consistency pins the core’s ecosystem check at 404ebb9, the pain001 0.0.88 release, which audits the five SDK repositories with the rest of the family; the job is now named “Every family publication agrees”. Regenerated from pain001’s scripts/render_ecosystem_workflows.py.

  • The README’s ecosystem table lists the fourteen suite repositories: the Go, Java, Rust, .NET and C++ SDKs joined the family at 0.0.88.

Security

  • The docs toolchain requires soupsieve 4.0 or newer: earlier versions have a polynomial-time denial of service in the descendant combinator (GHSA-mmhr-3c5m-f64f). Only the hash-pinned docs set contained it, so the published package is unaffected. The docs set now compiles from Python 3.11.5, the oldest version soupsieve 4 supports.

[0.0.87] - 2026-11-12

Changed

  • Release 0.0.87 in lockstep with the pain001 suite.

  • Requires pain001 0.0.87 (floor >=0.0.87,<1), the suite version.

  • Suite Consistency no longer runs on another workflow’s completion (workflow_run). It audits the family daily and on dispatch, as the core’s does, since that trigger runs with the default branch’s trust whatever started it. Regenerated from pain001’s scripts/render_ecosystem_workflows.py.

Added

Fixed

  • Suite Consistency pins the core’s ecosystem check at febcd0a, which accepts the five pain001 SDK repositories as known members not yet released. From the 0.0.86 release until this change every scheduled run failed on them as undeclared repositories.

[0.0.86] - 2026-11-11

Added

  • Add a code of conduct (adapted from Contributor Covenant 2.1, the suite’s shared text) and a markdownlint CI job (docs-lint.yml, configured in .markdownlint-cli2.jsonc); fix the Markdown it flagged.

  • Run every Python block in the README as part of the test suite, so the Quick Start cannot drift from the helpers it calls.

  • Add DEVELOPMENT.md, mapping every CI job to the local command that reproduces it, and a devcontainer that installs the project and runs make check.

  • Spell-check the tree with codespell in the docs-lint workflow, the pre-commit hooks and make check (make spelling, configured in pyproject.toml).

  • A Docs check workflow (docs-check.yml) builds the manual with every Sphinx warning an error and runs Sphinx’s link check on every push and pull request (make docs, make linkcheck); the deploy build in docs.yml uses the same warnings-as-errors flags. Links into this repository’s main are skipped by the link check and checked offline instead, for README.md, ARCHITECTURE.md and DEVELOPMENT.md, by tests/test_readme.py.

  • The manual includes ARCHITECTURE.md and DEVELOPMENT.md, and its API reference documents the faces helpers (faces_list, faces_project).

  • KEYS.asc: the maintainer’s SSH signing keys in allowed-signers format, the same keys the core, MCP server and mockbank publish, so a signed commit or tag can be checked with git config gpg.ssh.allowedSignersFile KEYS.asc.

Changed

  • Release 0.0.86 in lockstep with the pain001 suite and require core pain001 0.0.86; the locks resolve its published artifacts.

  • Align the pre-commit ruff and black hooks with the locked versions, drop the ruff-format hook that fought black, and keep black off the vendored tests/test_suite_conformance.py, which the old hooks rewrote.

  • CONTRIBUTING.md points at make check and DEVELOPMENT.md, asks for the DCO sign-off, and states the dual Apache-2.0 OR MIT licence.

  • README: the ecosystem table lists all nine suite repositories, the Examples table lists 05_faces_commands.py, and Documentation links the manual’s API reference and DEVELOPMENT.md; tests/test_readme.py now checks both tables’ sources and the four documentation links.

  • ARCHITECTURE.md described pygls 1.x (pygls >=1.3,<2) and a six-helper, single-module server; it now states the pygls 2.x range the package requires, maps every module, the formatting, symbol and workspace commands, and links ADR 0003 on explicit handler registration.

  • The record walkers behind diagnostics, the quick fix and the outline (_record_line_offsets, _record_close_positions) share one brace scan, _braces, which skips strings with a regular expression. Both are within every complexity ceiling and leave the baseline; their output is unchanged on 300,000 random documents and pinned by tests/test_record_scanning.py.

  • server.py (976 lines) is split under the 500-line ceiling: the schema loader (_schema.py), the record checks and their lsprotocol mapping (_diagnostics.py), the record walkers (_records.py), formatting and the outline (_layout.py), the corpus commands (_corpus.py) and compute_diagnostics_csv (into _csv.py) move out, and server.py imports them back, so every name it exposed is still importable from pain001_lsp.server. complexity-baseline.json is now empty.

  • The README demo is served from raw.githubusercontent.com on main instead of a relative path, which PyPI does not resolve; tests/test_readme.py fails on any relative README image.

Removed

  • CODE-OF-CONDUCT.md, a second copy of the code of conduct; CODE_OF_CONDUCT.md is the one GitHub and GOVERNANCE.md link to.

[0.0.85] - 2026-11-10

Changed

  • Release 0.0.85 in lockstep with the pain001 suite and require core pain001 0.0.85; the locks resolve its published artifacts.

  • Point usage questions at this repository’s GitHub issues: SUPPORT.md, SECURITY.md, GOVERNANCE.md, CODEOWNERS and the issue-template contact link no longer send them to GitHub Discussions, which the suite does not use.

[0.0.84] - 2026-11-09

Added

  • Audit all nine suite repositories and their publication channels daily and after publishing workflows, using the shared checker pinned to a core commit.

Changed

  • Release 0.0.84 with the suite and require core pain001 0.0.84. Regenerate dependency locks from the published core artifacts.

Fixed

  • Include the declared licence files and README in the container build context before package installation.

  • Emit the existing Apache-2.0 OR MIT licence as a PEP 639 expression. Remove the erroneous proprietary classifier produced by legacy Poetry metadata.

[0.0.83] - 2026-11-09

The pain001 family’s 0.0.83 release, on the pain001 0.0.83 floor (pyproject.toml and requirements/runtime.in). Nothing in the language server changed beyond the version and the floor: the core functions it calls (pain001.lsp.diagnostics, validate_iban, validate_bic) did not change in pain001 0.0.83, whose work is XML signing and signature verification.

Changed

  • Requires pain001 >= 0.0.83, < 1; poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.83.

  • Build and audit tools: virtualenv 21.15.6, dulwich 1.2.21, anyio 4.16.0 and pip 26.3 (Dependabot #95, #97, #98 and #99).

  • CI actions: github/codeql-action 4.39.3, docker/login-action 4.8.0, sigstore/cosign-installer 4.1.3 and softprops/action-gh-release 3.0.4 (Dependabot #102).

[0.0.82] - 2026-11-09

The pain001 family’s 0.0.82 release, on the pain001 0.0.82 floor (pyproject.toml and requirements/runtime.in). Nothing in the language server changed beyond the version and the floor: the core functions it calls (pain001.lsp.diagnostics, validate_iban, validate_bic) behave the same in pain001 0.0.82.

Changed

  • Requires pain001 >= 0.0.82, < 1; poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.82.

  • Development tools: ruff 0.17.4 and filelock 4.2.9 (Dependabot #93 and #94).

[0.0.81] - 2026-11-08

The pain001 family’s 0.0.81 release, on the pain001 0.0.81 floor (pyproject.toml and requirements/runtime.in). Nothing in the language server changed beyond the version and the floor: the core functions it calls (pain001.lsp.diagnostics, validate_iban, validate_bic) are the same in pain001 0.0.81.

Changed

  • Requires pain001 >= 0.0.81, < 1; poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.81.

0.0.80 - 2026-11-07

The pain001 family’s 0.0.80 release, on the pain001 0.0.80 floor (pyproject.toml and requirements/runtime.in). pain001 0.0.80 is a security release of the core’s REST API (GHSA-vv3x-jpvc-w468). Nothing in the language server changed beyond the version and the floor.

Changed

  • Requires pain001 >= 0.0.80, < 1; poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.80.

0.0.79 - 2026-11-07

The pain001 family’s 0.0.79 release, on the pain001 0.0.79 floor (pyproject.toml and requirements/runtime.in).

Fixed

  • Only payment documents get diagnostics. The VS Code extension sends every JSON file to the server, and the server read every non-CSV document as payment records, so package.json and tsconfig.json were told “‘id’ is a required property”. A JSON document is now diagnosed only when it is an array, or an object with at least one field of the active message type’s schema as a key; text that does not parse yet counts when it opens an array or quotes one of those fields as a key. Any other document is published with no diagnostics, which clears what an earlier version left. compute_diagnostics itself is unchanged. The check is is_payment_document in pain001_lsp._documents, where is_csv_uri (formerly server._is_csv_uri) also moved.

  • The README still called the faces commands unreleased; they shipped in 0.0.78.

Added

  • VS Code: a pain001.filePattern setting, the glob of JSON files sent to the server. The default, **/*.json, keeps the files reached before; set it to, for example, **/*.pain.json to stop completion, hover and formatting in other JSON files too.

Changed

  • Requires pain001 >= 0.0.79, < 1; poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.79.

  • The release page is written by the release workflow in the suite’s shape. A last release-notes job, once the SBOMs and the .vsix are attached, runs scripts/release_notes.py publish: the title is pain001-lsp X.Y.Z, the Highlights come from releases/vX.Y.Z.md, What’s Changed lists the pull requests merged since the previous release, and Checksums gives the SHA-256 of every asset. The 0.0.77 and 0.0.78 pages had to be rewritten by hand.

  • Release notes live in releases/vX.Y.Z.md, starting with releases/v0.0.78.md (the published 0.0.78 Highlights). tests/test_release_notes.py fails a pull request whose CHANGELOG dates the version while that note is missing or has other than 2 to 4 highlights. Both files are vendored byte-identical from pain001.

0.0.78 - 2026-11-07

The pain001 family’s 0.0.78 release, on the pain001 0.0.78 floor (pyproject.toml and requirements/runtime.in), which adds faces.

Added

  • Faces commands (pain001 >= 0.0.78): pain001.faces.list names each face, its target and the binding keys it accepts; pain001.faces.project takes a pain.001 document as XML text (the editor’s buffer) with a face, an optional version and optional bindings, and returns the payload for UK Open Banking v4.0.1, Berlin Group NextGenPSD2 v1.3.16 or ISO 20022 element names with a loss report of every source element not carried. A refusal, a file path instead of XML or an entity declaration is an {"error": ...} payload; with an older pain001 both commands say which version they need. The pure helpers are faces_list and faces_project in pain001_lsp._faces.

  • examples/05_faces_commands.py, run by make examples and CI.

Changed

  • The executeCommand argument helper moved to pain001_lsp/_commands.py, shared by the corpus and faces commands; server.py shrinks from 983 to 979 lines.

0.0.77 - 2026-11-06

The pain001 family’s 0.0.77 release, on the pain001 0.0.77 floor. 0.0.76 was prepared on 2026-11-05 but never tagged or published, so its changes ship here.

Security

  • The VS Code extension overrides minimatch to ^11.0.1 (GHSA-g3hv-75wg-m249): vscode-languageclient 9.0.1 resolved 5.1.9.

  • requirements/build.txt floors dulwich at 1.2.20 and virtualenv at 21.15.1.

Changed

  • Requires pain001 >= 0.0.77, < 1 (pyproject.toml and requirements/runtime.in); poetry.lock and the hash-pinned runtime, ci, docs and fuzz sets move pain001 to 0.0.77.

  • Development tools: ruff 0.17.3, complexipy 8.3.0, pytest 9.2.0, pytest-benchmark 5.3.1, mypy 2.4.1 and mutmut 3.8.1.

  • CI: one stable required check per versioned matrix.

0.0.75 - 2026-10-27

The pain001 family’s 0.0.75 release, on the pain001 0.0.75 floor, and the first on Python 3.11 or newer.

Security

  • No advised filelock installs from any requirements tree (GHSA-9vhv-jxww-xfm5, GHSA-567q-5qvw-pf92, GHSA-v4x8-3wr5-3wq8; audit.txt, build.txt and ci.txt pinned 4.1.0 on Python 3.10, which pip-audit on 3.12 skips). filelock 4.2.1 needs Python 3.11: build.txt and audit.txt, installed only on 3.12, are compiled for 3.12 with filelock >= 4.2.1; ci.in installs pip-audit, which pulls filelock in, only on 3.11+.

  • build.in floors virtualenv >= 21.14.13 (GHSA-cx6f-vjhg-g3cc; was 21.14.12) and dulwich >= 1.2.19 (was 1.2.17), in place of Dependabot #67 and #68, whose recompiles dropped the per-Python markers.

  • The VS Code extension’s dev tooling overrides @azure/msal-node to 7.1.0 (GHSA-xcx6-jx8p-x3h5; the lock held 6.0.1). @vscode/vsce pulls it in through @azure/identity, whose every release still requires msal-node 6.x, and all of 6.x is affected; the extension’s publish step signs in with VSCE_PAT, not an Azure credential.

  • certifi >= 2026.10.23 in the audit, build, CI and docs requirement sets (Dependabot #73 and #74, recompiled with each file’s own command so the per-Python markers stay).

  • The VS Code extension’s dev lockfile takes qs 6.17.0 (GHSA-6pq8-jhm2-2vcw; the lock held 6.16.0), Dependabot #76, which arrived after the release PR merged.

Removed

  • Python 3.10 support. The floor is 3.11 (python = ">=3.11,<4.0"), the pain001 suite’s decision for 0.0.75: 3.10 reaches end of life in October 2026, and current hypothesis releases need 3.11, which broke Dependabot’s resolver for siblings on a 3.10 floor. CI tests 3.11 to 3.14; ruff, black and mypy target 3.11; the tomli backport and the 3.10 branches in scripts/check_suite_consistency.py and the tests are gone, and poetry.lock and the 3.11 requirement sets are re-locked (only the packages for 3.10 leave). pip keeps 3.10 users on pain001-lsp 0.0.74.

Changed

  • Dependabot’s weekly pip run splits by Python. It resolved the “/” job on Python 3.10, then the floor in pyproject.toml, and since requirements/audit.in and build.in floor filelock >= 4.2.1 (Python 3.11+), any update touching them would end in dependency_file_not_resolvable, as it already did in camt053-mcp and acmt001-mcp. “/” now leaves the 3.12 sets (audit, build, fuzz) to a second pip entry for /requirements, which runs on Dependabot’s newest Python.

0.0.74 - 2026-10-21

The pain001 family’s 0.0.74 release, on the pain001 0.0.74 floor. The family skips 0.0.73, which pain001-mcp published on its own.

Fixed

  • A document json.loads rejects without a JSONDecodeError no longer raises out of the handlers. Nesting deeper than the recursion limit (about ten thousand open brackets) and an integer longer than 4300 digits used to leave the editor with stale diagnostics; diagnostics now report one Invalid JSON: ... error at the top of the document, and formatting, the outline and the quick fix offer nothing, as they do for a syntax error.

  • Completion names pain.001.001.13 (Customer Credit Transfer Initiation V13) and pain.008.001.08 (Customer Direct Debit Initiation V08) instead of showing the bare id, and the VS Code extension’s pain001.messageType setting lists both.

  • The GitHub release is titled pain001-lsp X.Y.Z, and the VS Code extension job waits for the release before attaching the .vsix (v0.0.72 was published without it).

  • The scheduled suite-consistency check compares this tree with pain001-lsp’s own PyPI release, not the core’s.

  • The Docker image installed pain001>=0.0.52,<0.0.54 before this package by default and was labelled Apache-2.0 only.

Changed

  • pain001 floor raised to >=0.0.74,<1. The 0.0.72 entry below says the floor moved to 0.0.72, but pyproject.toml still said >=0.0.71.

  • Every pip install in CI, the release job and the Dockerfile installs from hash-pinned requirements/*.txt, through scripts/ci-install.sh in CI; the base image is pinned by digest. OpenSSF Scorecard’s Pinned-Dependencies check goes from 2 to 10.

  • The VS Code extension is packaged with @vscode/vsce 4 on Node.js 22, which drops braces (GHSA-vfj7-8cjw-p6xm, no fixed release) and markdown-it (GHSA-gfr2-rv44-r23c, GHSA-vfgh-jcmw-g37q) from its lockfile.

  • Hash-pinned build, docs and audit sets floor certifi at 2026.10.19 (GHSA-6q48-6c7w-9gw2), urllib3 at 2.8.0 and virtualenv at 21.14.8.

  • The README follows the portfolio template, with a VHS demo, and the manual adds docs/COMPARISON.md, docs/BENCHMARKS.md and docs/POLICIES.md; it builds without warnings and links the author on every page.

  • CITATION.cff lists the licence as Apache-2.0 and MIT, the form CFF 1.2.0 accepts; Apache-2.0 OR MIT failed schema validation.

Added

  • scripts/check_complexity.py holds every function to the portfolio ceilings against complexity-baseline.json, which may only shrink: three offenders, down from four, after compute_diagnostics was split into helpers within every ceiling.

  • fuzz/fuzz_documents.py, an Atheris harness over every document handler, run on each change and weekly.

  • The Security workflow audits the poetry.lock environment and every requirements/*.txt with pip-audit.

  • scripts/verify_versions.py checks every version source and fails when the pain001 floor differs from the release version; the “Version sources agree” workflow runs it on each pull request.

  • ruff and black cover scripts/, fuzz/ and examples/.

[0.0.72] - 2026-09-28

Aligns on pain001 0.0.72: ISO 20022 schemas and Schematron validation rules, enhanced CI provenance attestations, and dependency updates.

Changed

  • pain001 floor raised to >=0.0.72.

  • Version aligned to 0.0.72 across all pain001 suite packages.

  • The README and the documentation index link What is pain.001?, an explainer of the message’s structure, versions and a schema-valid example.

[0.0.71] - 2026-09-26

Aligns on pain001 0.0.71: request-local CEL policy rules, review-only correction suggestions, explicit SFTP delivery, and release artifact inspection.

Changed

  • pain001 floor raised to >=0.0.71.

  • Version aligned to 0.0.71 across all five pain001 packages.

[0.0.70] - 2026-09-13

Aligns on pain001 0.0.70: importing the core no longer loads its CLI, which makes the server start faster; the rail rules read the CSV pipeline’s column names as well as the short ones.

Changed

  • pain001 floor raised to >=0.0.70.

  • Version aligned to 0.0.70 across all five pain001 packages.

[0.0.69] - 2026-09-12

Aligns on pain001 0.0.69, the third example-corpus release: the tier 2 and 3 market packs reach pain001.corpus.list unchanged.

Changed

  • pain001 floor raised to >=0.0.69.

  • Version aligned to 0.0.69 across all five pain001 packages.

[0.0.68] - 2026-09-12

Aligns on pain001 0.0.68, the second example-corpus release: the pain001.corpus.list command now sees the tier-1 market packs and the coverage files named by recipe and focus.

Changed

  • pain001 floor raised to >=0.0.68.

  • Version aligned to 0.0.68 across all five pain001 packages.

[0.0.67] - 2026-09-12

The example corpus reaches the editor: the validated pain.001 files pain001 0.0.67 ships, with their provenance, are listable and readable through workspace commands.

Added

  • Example corpus commands. workspace/executeCommand gains pain001.corpus.list (the validated example files pain001 ships: market scenarios per country and rail with bank variants, and the schema coverage sets, filterable by kind, country and message type) and pain001.corpus.get (one file’s XML with its provenance sidecar), backed by the pure helpers corpus_list and corpus_get. They delegate to pain001.corpus, which shipped in pain001 0.0.67; the dependency floor is now 0.0.67, and with an older library they still return an error payload rather than failing at import.

Changed

  • pain001 floor raised to >=0.0.67, the first release with pain001.corpus.

  • Version aligned to 0.0.67 across all five pain001 packages.

  • Dev tools: ruff 0.16.6, packaging 26.3, pytest-benchmark 5.3.0; GitHub Actions group bumped; the vendored conformance file refreshed to the canonical copy.

[0.0.66] - 2026-09-12

Aligns the pain001 suite on 0.0.66, the core release that adds cross-record duplicate detection, the /api/v1/ui dashboard, GPG flags and OpenTelemetry spans. Nothing in this package changed.

Changed

  • Version aligned to 0.0.66 across all five pain001 packages.

[0.0.65] - 2026-08-29

Aligns the pain001 suite on one version number, and adds the scheduled drift check this repository was missing.

Added

  • scripts/check_suite_consistency.py and a scheduled Suite Consistency workflow compare this tree, and every published member of the suite, against PyPI. A member left a release behind still installs and still passes its own tests; only the index disagrees, and only if somebody looks.

Changed

  • Version aligned to 0.0.65 across all five pain001 packages, which had drifted to 0.0.62, 0.0.64, 0.0.63, 0.0.62 and 0.0.63.

  • Refreshed the vendored tests/test_suite_conformance.py to the current canonical copy. Its drift-check probe now matches the bare word pypi rather than the hostname, which CodeQL read as an incomplete URL sanitisation.

[0.0.64] - 2026-08-29

Brings this repository onto the suite conformance gate.

Added

  • benches/bench_diagnostics.py — how long the editor waits.

    A language server is not measured in throughput. It is measured against a person typing: the client recomputes diagnostics on every change, so the only question is whether the answer arrives before the user notices. The table marks two thresholds — 16 ms (one frame, feels instantaneous) and 100 ms (the limit of “reacting instantly”).

    The number to read is the budget: how large a document stays interactive. It is a budget, not a score.

    For comparison across the three language servers in the suite, all doing the same shape of work:

    Server

    Cost

    Interactive budget

    bankstatementparser-lsp

    0.0018 ms/entry

    ~56,000 entries

    pain001-lsp

    0.08 ms/record

    ~1,200 records

    camt053-lsp

    0.86 ms/record

    ~116 records

    The malformed column matters too: that is the state a document spends most of its life in while somebody is typing. A linter fast on valid input and slow on invalid input is slowest exactly when the editor calls it most.

    Nothing asserts a timing threshold — wall-clock is not comparable between machines, and a flaky performance gate teaches people to ignore red. CI runs --quick, so a benchmark that stops compiling fails the build rather than rotting.

    This server sits comfortably inside the interactive band — 0.08 ms per record, a budget of roughly 1,200 records — and returns immediately on malformed input rather than doing partial work.

  • tests/test_suite_conformance.py — invariants shared by every repository in the suite, vendored from one canonical copy and checksummed by its own test.

Changed

  • CI lints, formats and runs benches/ alongside everything else.

  • tests/test_suite_conformance.py is excluded from black: it is generated, and the suite uses three different line lengths.

[0.0.63] - 2026-08-28

Changed

  • A plain pytest now enforces the same 100% coverage floor CI does. --cov, --cov-branch and --cov-report=term-missing move into [tool.pytest.ini_options] addopts. Before this, running pytest locally measured no coverage at all: the developer saw green and learned otherwise from the build.

    The gate itself was never missing — CI has always run --cov-fail-under=100, so nothing could regress onto main unnoticed. What was missing was the local half of it.

  • The floor is stated once. It was in two places, the --cov-fail-under flag in ci.yml and fail_under under [tool.coverage.report], which is how two numbers meant to be equal stop being equal. pytest-cov honours fail_under from the coverage config, so the flag is gone from both addopts and the workflow and the number lives in exactly one place.

    CI now runs pytest tests/ --cov-report=xml -v; only the XML report is CI-specific.

Verification

Coverage was already 100% and remains so. The gate was mutation-tested: adding a function no test calls drops coverage below the floor and fails the run, locally as well as in CI.

[0.0.62] - 2026-08-21

Performance

  • The bundled JSON Schema is cached. Every interactive entry point goes through _load_schema, which re-read and re-parsed it from disk on each call — a file read and a JSON parse per keystroke, to answer from data that ships read-only inside the package and cannot change at runtime.

    call

    before

    after

    completion_items

    0.0805ms

    0.0029ms (28x)

    hover_text

    0.0794ms

    0.0002ms (478x)

    missing_required_fields

    0.0816ms

    0.0004ms (196x)

    Not user-visible — 0.08ms is imperceptible — so this is waste removed rather than latency fixed. It surfaced from benchmarking: three functions measuring identically to three decimal places were not three measurements, they were the same file read three times.

Added

  • Benchmarks for completion, hover and the missing-field check, the per-keystroke paths the 0.0.61 set did not cover. The guard is a property rather than a threshold — at microsecond scale any wall-clock bound loose enough for CI would not notice the cache being removed, so the test asserts the schema is not re-read across 150 interactive calls.

  • A locked CI job that installs from poetry.lock and runs the suite against the pinned set. 0.0.61 added poetry check --lock after finding the lock silently stale, but nothing installed from it — so it proved the lock matched pyproject.toml and nothing about whether the pinned versions worked. The pip-based jobs still install against the live range; the two answer different questions.

[0.0.61] - 2026-08-20

Suite release with pain001 0.0.61. No change in this package; the improvement comes from the core.

Performance

  • Diagnostics are 3-6x faster, because pain001 0.0.61 fixed a quadratic in pain001.lsp.diagnostics. Each cell’s character span was computed by re-splitting the line and summing the lengths of every preceding cell, once per cell — and eagerly, before knowing whether there was anything to report, so a clean document paid the whole cost to produce no diagnostics. That is the common case in an editor, on every keystroke.

    rows

    before

    after

    500

    43.4ms

    7.6ms (5.7x)

    2000

    102.0ms

    32.1ms (3.2x)

    Diagnostics output is unchanged. This package’s benchmark had recorded the pre-fix behaviour as “sublinear (~2.4x)”, which was the quadratic per-row cost swamping the row count rather than a fixed overhead; the ratio is now 4.23x for 4x the rows, which is what linear looks like. The benchmark’s documentation is corrected accordingly.

[0.0.60] - 2026-08-20

Lockstep release with pain001 0.0.60. No functional change in this package.

pain001-lsp is a lockstep member of the suite: its version tracks the core so an editor integration can pair pain001-lsp==X with pain001==X without consulting a table. pain001.suite records which members follow that rule, and the core enforces it daily.

The pain001>=0.0.55 floor is deliberately unchanged, which is worth stating because the core’s 0.0.60 does contain a pygls 1.x -> 2.x migration and this package requires pygls>=2.1,<3. That migration is confined to pain001.lsp.server, the core’s own language-server entry point. This package does not import it — it imports pain001.lsp.diagnostics, which has no pygls dependency at all, plus pain001.validation and pain001.constants. pain001.lsp.__init__ re-exports only from diagnostics, so importing the package does not pull the migrated module in either. Checked rather than assumed, because a floor that is too low here would surface as an import error in a user’s editor rather than at install time.

[0.0.59] - 2026-08-20

Rejoins lockstep with pain001. This package sat at 0.0.54 while the core reached 0.0.59 — five releases of silent drift, which is what pain001’s suite-consistency check now reports rather than leaving a user to notice.

The version jumps 0.0.54 -> 0.0.59 with no intervening releases. The numbers in between were never published for this package; they exist so that a user reading pain001==0.0.59 and pain001-lsp==0.0.59 can trust the two belong together, which is the point of lockstep versioning.

Changed

  • Migrated to pygls 2.x (#13).

  • Dependency updates: mypy 2.3.0, ruff 0.16.2, and the grouped GitHub Actions bump (#9, #10, #12).

  • Dependabot config repaired — an invalid key meant github-actions updates were never grouped (#8).

[0.0.54] - 2026-07-18

Changed

  • Require pain001 >= 0.0.55 to pick up the upstream security fix shipped in that release (security-fix propagation across the pain001 suite). No LSP-surface behaviour changes.

[0.0.53] - 2026-06-19

Added

Two new editor features that surface previously-missing IDE ergonomics for pain.001 payment-data JSON files:

  • textDocument/formatting - pretty-print the document as a two-space-indented JSON array with a trailing newline. The handler re-serialises via json.dumps(parsed, indent=2), leaves malformed JSON untouched (diagnostics already flag the syntax error), and returns no edits when the document is already formatted (so the editor’s “format on save” stays idempotent).

  • textDocument/documentSymbol - return one DocumentSymbol per top-level record so editors can populate the outline pane, jump to a specific payment, and code-fold individual records. Each symbol uses the record’s id field as its name (falling back to <record N>) and payment_id as its detail; the range spans the record’s opening { through its closing }.

Changed

  • Pinned to pain001 >= 0.0.53 so the new public-API symbols (sanitize_to_charset, the SEPA B2B profile) are available.

  • Mypy: added an ignore_missing_imports override for the jsonschema.* module so poetry run mypy pain001_lsp is clean out of the box (no types-jsonschema install needed).

Quality gates

  • pytest: 81 tests, 100% line + branch coverage (was 79, +2 new feature handlers exercised by 13 new tests).

  • interrogate: 100% docstring coverage.

  • ruff + mypy all clean.

0.0.52 - 2026-06-18

Added

  • Initial release of pain001-lsp, a pygls-based Language Server Protocol (LSP) server for authoring pain001 payment-data JSON files (Python 3.10+)

  • A pain001-lsp console entry point that starts the language server over stdio for editor LSP clients

  • Diagnostics - schema validation of each record against a message type’s input JSON Schema (missing required fields, types, patterns) plus IBAN / BIC validation of identifier fields

  • Completion - every input field (with its schema description) and the full list of supported pain message types

  • Hover - schema descriptions for the field under the cursor

  • Code actions - multi-record “Add missing required fields” quick-fix backed by missing_required_fields(...) and build_insert_text(...); uses the cursor’s line to pick the target record (falls back to the first one) and inserts JSON placeholder lines ("" for strings, 0 for numbers, false for booleans) before that record’s closing brace.

  • Workspace configuration - editors override the default message type either at startup (initializationOptions: {"messageType": ...}) or live via workspace/didChangeConfiguration (accepting either a nested {"pain001": {"messageType": ...}} payload or a flat {"messageType": ...} payload).

  • Pure, importable helper functions (compute_diagnostics, completion_items, hover_text, missing_required_fields, build_insert_text) backed by the pain001 public API, so editor behaviour matches the CLI, REST API, and MCP server

  • Three runnable examples (examples/01_lsp_helpers.py, examples/02_quick_fix.py, examples/03_configure_message_type.py)

  • Part of the pain001 suite alongside the core pain001 library and the pain001-mcp Model Context Protocol server

  • Versioning aligned with pain001 and pain001-mcp: the three packages in the suite ship under matching release numbers

  • Quality gates pinned at 100% from the initial release:

    • pytest --cov=pain001_lsp --cov-branch --cov-fail-under=100 (68 tests exercising every line, branch, and LSP handler in pain001_lsp/server.py, including the multi-record code action, workspace configuration paths, brace-walker fall-through, and an end-to-end stdio handshake against a real subprocess)

    • interrogate --fail-under=100 for module and function docstring coverage

    • Every example script is also exercised by pytest so breakage is caught at the test-suite level

  • Quality workflows - ci.yml enforces ruff, mypy, the 100% pytest coverage gate, the 100% docstring gate, and the example scripts on Python 3.10/3.11/3.12; security.yml runs bandit + pip-audit; codeql.yml runs GitHub’s CodeQL Python analysis weekly.

  • Security policy (SECURITY.md) describing the (small) threat model: stdio-only, bounded configuration, no caller-supplied filesystem paths.

  • scripts/verify_versions.py - pre-release script asserting __version__, pyproject.toml, and CHANGELOG.md agree.