xbsl 0.21.0

Инструментарий XBSL (1С:Элемент): линтер, LSP, документация, индекс проекта, скаффолдинг метаданных

xbsl

English · Русский

The XBSL (1C:Element) toolkit: a linter with autofixes, an LSP server, a project index, platform documentation search, metadata scaffolding and an MCP server for AI agents. It works on Name.yaml (element description) and Name.xbsl (code module) pairs – before the server-side compilation that happens on deploy.

Before 0.16 the project was named xbsl-lint (the xbsllint package). The old names keep working: the xbsllint* commands are aliases of the new ones, import xbsllint returns the xbsl modules, and both spellings of the environment variables and entry-point groups are honored.

Not affiliated with 1C. "1C:Element", "1C:Fresh" and related names are trademarks of their respective owners. Language data is generated from your own distribution. See NOTICE.

Development notes and updates (in Russian): the 1C × AI: engineering workshop Telegram channel.

Why

1C:Element has no external tooling: the only code check is the server-side compilation on deploy – it is slow and knows nothing about project conventions. xbsl gives fast local feedback, catches what the compiler does not check at all, and takes over the metadata mechanics – creating objects, attributes and forms.

How it works

One engine, four surfaces. The core reads the Имя.yaml + Имя.xbsl pairs, and the scaffolding writes them back; the CLI, the LSP server, the MCP server and the web UI are thin adapters over the same core, so every surface sees the same rules, data and templates:

Step 1: generate the language data

The linter relies on language tables (bilingual keywords, operators), an stdlib type catalog, and the configuration metamodel (element properties). XBSL is built on Eclipse Xtext + ANTLR; these are extracted from your 1C:Element distribution (the InternalBsl.g grammar, the documentation, and the .xcore metamodel) and are NOT bundled in this repository. Generate them locally:

python tools/extract_grammar.py   --dist "<path to the 1C:Element distribution>"
python tools/extract_stdlib.py    --dist "<path to the 1C:Element distribution>"
python tools/extract_metamodel.py --dist "<path to the 1C:Element distribution>"

The scripts auto-detect the platform version and place the data under xbsl/data/element/<version>/ (this folder is gitignored). Without the data, the linter and the tests will tell you to generate it. Pass --data-dir (or set XBSL_DATA_DIR) to write the data somewhere else – for instance into a private package that ships it, see Extending.

Step 2: install and run

pip install xbsl            # or, from a clone: pip install -e .
xbsl path/to/sources        # or: python -m xbsl path/to/sources
xbsl self-update            # upgrade to the latest PyPI version

self-update upgrades the package by unpacking the wheel straight into site-packages – safe even when pip install --upgrade fails with WinError 32 because an exe is busy (the typical case: xbsl-lsp.exe held by the VS Code LSP server, xbsl-mcp.exe by an agent's MCP session). The busy stubs are left alone and pick up the new code on the next start; restart the long-living processes after the update. --version X.Y.Z installs a specific version. In an editable install from a clone the command refuses – git pull updates that one.

The extractors from step 1 ship with the repository, not with the PyPI package – clone the repository to generate the data.

The hot modules (the lexer and the parser) can be compiled by mypyc into C extensions: XBSL_MYPYC=1 at build time (needs mypy and a C compiler: MSVC Build Tools on Windows, Xcode CLT on macOS, gcc on Linux). Users never need a compiler: the ready-made native wheels are built by CI (native-wheels.yml), and without a matching wheel the package runs as plain Python – no compiler, no loss of functionality.

Flags: --list-rules, --where (data root, source and versions), --select/--enable/--ignore (by rule id, rule group – the part of the id before / – or tier letter), --fix, --baseline/--write-baseline, --element-version, --data-dir, --lang, --format text|json|codeclimate. --fix repairs the mechanical findings in place – trailing whitespace, typography characters (em dash → en dash, … → ..., curly quotes and comment guillemets → straight), and mixed newlines (normalized to the dominant style) – then reports whatever is left. It only applies unambiguous edits and only for rules active in the run (so --fix --enable typography also pays down the em-dash/guillemets debt); anything needing judgment is never touched. For editor integration, --stdin --filename NAME checks a single buffer read from stdin (per-file rules only); the JSON payload ({diagnostics, summary}) is the same one the MCP server returns. xbsl --index PATH dumps a JSON index of the project to stdout instead of linting – the objects (with tabular sections, module-declared local types and the member families for dot completion), the method declarations with their annotations and the named form components, with POSIX paths relative to the root and 1-based lines – for go-to-definition and completion in editors. --format codeclimate emits a GitLab Code Quality report (Code Climate issues) with paths relative to the current directory – run it from the repository root and save the output as the codequality artifact.

Output language

Rule titles and diagnostic messages come in Russian and English. The language is picked by --lang ru|en > the XBSL_LANG env var > the system locale > Russian. Type names, keywords and other XBSL text inside a message are never translated – only the wording around them. The MCP server and the web panel follow the same setting (the web panel also has an in-page RU/EN toggle).

Use in CI

xbsl exits non-zero only when a run produces an error-severity finding, so it works as a pipeline gate as-is – warnings and info do not fail the build. The one prerequisite is the language data (see Step 1): generate it in the job (the extractors ship with the repository, so check the repo out), or depend on a package that ships the data via the xbsl.data entry point (see Extending) and just pip install it.

GitHub Actions

lint:
  runs-on: ubuntu-latest
  steps:
    - uses: actions/checkout@v4
    - uses: actions/setup-python@v5
      with: { python-version: "3.12" }
    - run: pip install xbsl
    # generate the data from your 1C:Element distribution (or install a package that ships it):
    - run: |
        python tools/extract_grammar.py  --dist "$ELEMENT_DIST"
        python tools/extract_stdlib.py   --dist "$ELEMENT_DIST"
        python tools/extract_metamodel.py --dist "$ELEMENT_DIST"
    - run: xbsl e1c/          # fails the job on any error-severity finding

GitLab CI (Code Quality widget)

--format codeclimate writes a Code Climate report that GitLab renders inline on the merge request. Run it from the repository root and save the output as the codequality report. The command still returns non-zero on error-severity findings, so artifacts.when: always keeps the report even when the job gates the pipeline (drop the gate with a trailing || true if you want the widget only):

lint:
  script:
    - pip install xbsl
    - xbsl --format codeclimate e1c/ > gl-code-quality-report.json
  artifacts:
    when: always
    reports:
      codequality: gl-code-quality-report.json

Rule tiers

The full list of all 86 rules (severity, default state, scope, links to platform documentation sections) is in docs/RULES.md; at runtime – xbsl list-rules. Below is an overview by tier.

• A. Structure and YAML – .xbsl/.yaml pairing, schema validity, Ид as a UUID, Ид uniqueness, Имя matching the file name.
• B. Text and conventions – typography (en dash, straight quotes), encoding/BOM/newlines/trailing whitespace, indentation and line length.
• C. Code structure – balance of blocks and ;, brackets, unused local and loop variables, a structure reference field that must be обз, plus the platform's code style conventions (the style/ group, see below).
• D. Semantics – against platform data and the project itself: types, enumeration values, cross-file consistency (see below).

The type rules of tier D cover every type position in code (новый, как casts, annotations, signatures) and every Тип: value in yaml (unions А|Б|?, generics, nullable): the root must be a known type – stdlib, a project object or a module-declared local type – and a dotted chain rooted at a project object must stay within the family that object generates: the derived types extracted from the distribution docs (Ссылка, Объект, СоздатьОбъект, the automatic forms...), its tabular sections and module structures. Namespace-qualified references (Справочник.X.Ссылка) also check that the object exists under that kind, and the values of project enumerations are verified both in code and in yaml bindings.

The cross-file rules of tier D catch what the compiler reports late or not at all: a yaml handler missing from the paired module, a foreign-subsystem type used without an Импорт: entry, a dynamic list typed by the automatic list form that misses an attribute of its object, a cross-component Компоненты.X.Метод() call to a method without a visibility annotation, environment mismatches (@НаСервере called from a client handler without @ДоступноСКлиента, a client-only module used from an HTTP service), reserved names (Тип/type as a field or parameter, a component property named like a built-in one), methods that nothing references, and top-level yaml properties against the configuration metamodel. The query/ group parses Запрос{ ... } blocks and verifies the tables of ИЗ/СОЕДИНЕНИЕ against the project objects and their tabular sections; a block with constructs outside the supported subset (temporary tables, unions, subqueries) is skipped whole rather than guessed.

Detailed group descriptions - query/ (a composite type in IN with a subquery), project/ (project properties), naming/ (the naming standard, the [morph] extra) and style/ (code-writing conventions and their on/off policy) - live in docs/RULES.md.

Baseline: adopt a rule on a legacy codebase

To enable a rule over code that already violates it without drowning in legacy findings, freeze the current findings into a baseline and hold only new code to the rule:

xbsl e1c/app --enable style --write-baseline baseline.json   # freeze the debt once
xbsl e1c/app --enable style --baseline baseline.json         # only NEW findings surface

A finding's identity is (file, rule, message) with an allowed count, so moving a line keeps its finding suppressed while a genuinely new violation surfaces. The summary reports how many findings the baseline suppressed and how many of its entries are now stale (debt paid down) – a signal to rewrite the file. Paths are stored relative to the baseline file, so commit it at the repository root and run the linter from anywhere.

The same file also records point exclusions with their reasons: an entry's value is either a bare count or {"count": N, "reason": "..."} – the reason says why the code is right on purpose. Reasons are written by the "Exclude the finding" lightbulb action of the VS Code extension (or by hand); --write-baseline keeps the reasons of the identities that survive a rewrite. The LSP server accepts the same --baseline FILE flag, so exclusions disappear in editors too. The identity includes the message text: write and check the baseline under the same output language.

Metadata scaffolding

The toolkit takes over the metadata mechanics: UUIDs, indentation, precise yaml insertions, duplicate checks and section/kind compatibility. The same operations are exposed through the CLI (subcommands, JSON output), MCP (the meta_* tools for agents) and LSP (the xbsl/meta* custom requests that power the VS Code metadata tree).

33 kinds of project element are creatable – from Справочник and Документ to ВиртуальнаяТаблица (paired with its mandatory .xbql query), ЗапланированноеЗадание, contracts, rights and commands. Each kind carries what the docs make mandatory: the platform's own default scope (ВПодсистеме – widen it deliberately with --scope), a module stub for the handler the kind cannot live without, and a note for whatever the generator must not invent for you. Kinds whose content is drawn in the designer (ПанельОтчетов, ПроцессИнтеграции) are deliberately absent.

xbsl new-project . vendor App                        # Проект.yaml + Проект.xbsl + a subsystem
xbsl new-object vendor/App/Основное Справочник Товары
xbsl add-field vendor/App/Основное/Товары.yaml реквизит Цвет --type Строка
xbsl add-form . --name Товары                        # object + list forms, registered
xbsl add-form . --name Товары --forms list-cards     # list form as a card grid
xbsl new-object ... HttpСервис Каталог --routes "GET /, POST /, GET /{id}"
xbsl add-route  .../Каталог.yaml "DELETE /{id}"      # url template + handler stub
xbsl add-subsystem vendor/App Задачи
xbsl add-dependency . e1c CurrencyConverter 2.0      # library into the project's Библиотеки
xbsl rename-object . Товары Номенклатура             # rename files + update references
xbsl set-access . --name Товары --default РазрешеноАутентифицированным
xbsl object-info . --name Товары                     # fields, tabulars, forms, namespace
xbsl project-info .                                  # projects, subsystems, objects by kind

Forms are generated with real content: input fields per attribute (including the standard Наименование / Номер / Дата and hierarchy support), dynamic-list columns, tabular-section tables, a report form with parameters; the form is registered in the owner's Интерфейс section. --dry-run prints the changes (with full file texts) without writing – this is how the VS Code extension applies them through its own undo-friendly edits.

--forms list-cards builds the list form as a card grid instead of a table: a ПроизвольныйСписок whose КонтейнерСтрок is a matrix group, plus the row component СтрокаСписка<Имя>. The card takes a title, a photo (a ДвоичныйОбъект.Ссылка attribute switches it to ПроизвольнаяКарточка with the image above the caption) and up to three more fields, dates formatted; notes report what landed on the card and what did not. --card-min-width sets the grid column width (default 400, 250 with a photo) and --card-placeholder the image shown when the photo is empty.

add-dependency attaches a library – the Библиотеки section of Проект.yaml (Имя, Поставщик, Версия). The version is the library's release version: a release is issued in the control panel, and a build version with a suffix (1.0-42) is rejected. Different versions of one library within a project are not allowed, so attaching an already attached library updates the version of the existing entry. What is attached now – project-info (projects[].libraries). The vendor, name, version and the qualified type names of a library come from parsing its archive: elemctl inspect <file.xlib>.

set-access edits КонтрольДоступа.Разрешения in place, aware of what each kind allows: --default sets the ПоУмолчанию right, --permission Чтение=РазрешеноВсем an individual one (custom rights of a ПравоНаЭлемент included), --calc-by fills РасчетРазрешенийПо – mandatory for РазрешенияВычисляютсяДляКаждогоОбъекта. Wrong methods, rights a kind does not have, and per-object rights on a НаборКонстант are rejected; the computed-permission handlers stay yours to write (notes say which). object-info reports the current permissions and the kind's rights, project-info the ПоУмолчанию of every object – no section there means the platform applies РазрешеноАдминистраторам.

rename-object renames the object's files (including its forms and the СтрокаСписка<Имя> row component) and rewrites references context-aware across the whole project: yaml type/table/form keys, = bindings and .xbsl code. Attributes, components or dynamic-list fields that merely share the old name are left alone, and so are string literals (UI text); --new-presentation/--old-presentation update the Заголовок/Представление values of the object and its forms. The object's Ид is untouched, so the platform keeps the stored data.

Extending: your own rules, data and severities

Three entry point groups let a separate package extend the linter without forking it. This exists for teams whose rules or language data cannot be published: keep those in a private package that depends on xbsl.

# pyproject.toml of your package
dependencies = ["xbsl>=0.16"]

[project.entry-points."xbsl.rules"]
myproject = "myproject.rules"        # importing the module runs its @rule decorators

[project.entry-points."xbsl.data"]
myproject = "myproject:data_root"    # a path, or a callable returning one

[project.entry-points."xbsl.severity"]
myproject = "myproject:severity_overrides"   # {rule id: "error"|"warning"|"info"|"off"}

Packages that declared the groups under the pre-rename name (xbsllint.rules/xbsllint.data/ xbsllint.severity) keep working: the legacy groups are scanned after the new ones.

The severity dict (or a zero-argument callable returning one) raises or lowers the default level of any rule – built-in or plugin – for every run in this installation: a project may treat, say, style/abbreviation-case as a warning while the published default stays info. "off" removes a rule from the default set (an explicit --select/--enable still turns it on, at its base level).

Install the package and the CLI, the MCP server and the web UI all pick everything up – no flags, no config file. A failing entry point raises instead of warning: a linter that silently drops a rule stays green in CI and guarantees nothing; an override naming an unknown rule id or level raises for the same reason. XBSL_NO_PLUGINS=1 ignores every external package (built-in rules, bundled data and default severities only).

LSP server (experimental)

xbsl-lsp (the [lsp] extra: pip install "xbsl[lsp]") runs the linter as a long-living Language Server over stdio: live per-file diagnostics as you type, project-wide diagnostics on save, go to definition, completion and hover over a resident project index, and quick-fix code actions - without paying the interpreter start-up cost per call. Flags: --project-root (the sources root relative to the workspace folder), --select/--ignore/ --enable, --data-dir. Any LSP-capable editor (VS Code, Neovim, JetBrains) can spawn it.

Documentation (searching the Element reference)

tools/extract_docs.py extracts the Element reference from a distribution (the server-with-IDE .car) into a docs.sqlite next to the language data: the stdlib pages (a type, its methods, properties, parameters) with cleaned HTML, a full-text index (SQLite FTS5, from the standard library) and canonical links back to the primary source (https://1cmycloud.com/docs/help/..., taken from the distribution's sitemap.xml). Page images are stored alongside. The 1C reference is copyrighted, so the database is not shipped in the package – you generate it from your own distribution, like the language data (step 1).

python tools/extract_docs.py --dist "$ELEMENT_DIST"

The runtime API xbsl.docs (search, page, tree, for_symbol, asset) reads docs.sqlite; with no database the search is simply empty. It powers the MCP tools (below) and – later – the reference panel in the VS Code extension.

MCP server

A thin adapter over the same core: an agent (e.g. Claude Code) can call the checks as tools and receive structured diagnostics.

pip install -e ".[mcp]"
claude mcp add xbsl -- xbsl-mcp

Tools: lint_paths(paths), lint_source(filename, content), list_rules(); documentation search – docs_search(query), docs_page(id), docs_symbol(name) (needs the docs.sqlite database, see above); type_members(name) – the members of a stdlib type with the return-type roots of its methods in one compact answer (cheaper than a docs page when only the member list matters); metadata scaffolding – meta_new_project, meta_new_object, meta_add_field, meta_add_route, meta_add_form, meta_add_subsystem, meta_add_dependency, meta_rename_object (with a dry_run plan mode), meta_set_access, meta_object_info, meta_project_info. Every writing meta_* tool applies the changes and returns the lint of the written files in the same response – creation and validation in one round trip. The core and the CLI do not require mcp – it lives only in the [mcp] extra.

Web interface

A local page: point it at a project folder and see the diagnostics. Standard library only (no external dependencies), binds to 127.0.0.1 only.

xbsl-web            # then open http://127.0.0.1:8771/

Per-tier rule toggles, a data-version selector, severity/text filters, dark/light theme; clicking a diagnostic opens the file in VS Code (vscode://).

Editor support (VS Code)

A VS Code extension in editors/vscode gives .xbsl syntax highlighting, live diagnostics as you type (--stdin), workspace diagnostics on save (a full linter run in the background brings the project-scope rules into the editor), index-based go-to-definition and completion across the project (xbsl --index), a form preview with a properties panel, and a deploy button powered by elemctl. It is published on the Marketplace and Open VSX; its README covers settings, behavior and requirements. See also the companion XBSL Debug extension from the elemctl project.

Element versions

The data is versioned by platform version:

xbsl/data/element/
    index.json            # { available: [...], default: "<version>" }
    <version>/{language.json, stdlib.json, metamodel.json}

Pick a version with --element-version / the XBSL_ELEMENT_VERSION env var / the index default; --version shows what is available. Add a new version by re-running the extractors with a new --dist.

The data root itself is resolved in this order: --data-dir > XBSL_DATA_DIR > a root supplied by an installed xbsl.data entry point > xbsl/data/element inside the package.

Tests

pip install -e ".[dev]"
pytest

Data-dependent tests are skipped automatically when the data has not been generated.

License

MIT – see LICENSE. Trademarks and data provenance – NOTICE. How to add a rule – CONTRIBUTING.md.