cobol-parser 3.1.1__tar.gz → 3.3.0__tar.gz

cobol_parser-3.3.0/PKG-INFO

@@ -0,0 +1,301 @@
+Metadata-Version: 2.4
+Name: cobol-parser
+Version: 3.3.0
+Summary: A comprehensive COBOL parser for Python with IBM Enterprise COBOL focus
+Author-email: Samuel Dion <sam94dion@gmail.com>
+License: AGPL-3.0-or-later
+Project-URL: Homepage, https://github.com/Landparty/cobol-parser
+Project-URL: Repository, https://github.com/Landparty/cobol-parser
+Project-URL: Bug Tracker, https://github.com/Landparty/cobol-parser/issues
+Keywords: cobol,parser,mainframe,ibm,enterprise,legacy
+Classifier: License :: OSI Approved :: GNU Affero General Public License v3 or later (AGPLv3+)
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Classifier: Topic :: Software Development :: Compilers
+Classifier: Topic :: Text Processing :: Linguistic
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Provides-Extra: fast
+Requires-Dist: cython>=3.0; extra == "fast"
+Provides-Extra: jcl
+Requires-Dist: jcl-parser>=2.0.0; extra == "jcl"
+Provides-Extra: pl1
+Requires-Dist: landparty-pl1-parser>=2.0.0; extra == "pl1"
+Provides-Extra: hlasm
+Requires-Dist: hlasm-parser>=0.2.2; extra == "hlasm"
+Provides-Extra: mcp
+Requires-Dist: mcp<2,>=1.10; extra == "mcp"
+Provides-Extra: dev
+Requires-Dist: jcl-parser>=2.0.0; extra == "dev"
+Requires-Dist: mcp<2,>=1.10; extra == "dev"
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
+Requires-Dist: pytest-xdist>=3.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Requires-Dist: black>=23.0.0; extra == "dev"
+Requires-Dist: mypy>=1.5.0; extra == "dev"
+Requires-Dist: cython>=3.0; extra == "dev"
+Provides-Extra: docs
+Requires-Dist: sphinx>=7.0.0; extra == "docs"
+Requires-Dist: sphinx-rtd-theme>=1.3.0; extra == "docs"
+Provides-Extra: profiling
+Requires-Dist: psutil>=5.9; extra == "profiling"
+Dynamic: license-file
+
+# cobol-parser
+
+A COBOL parser for Python with an IBM Enterprise COBOL focus. It turns COBOL
+source into a traversable AST (JSON export, error recovery, copybook
+expansion with a source map back to the original lines), and builds static
+analyzers, test-data generators, a z/OS JCL front end and an MCP server for AI
+agents on top of it.
+
+458 of the 459 NIST COBOL-85 programs in `examples/cobol/` parse cleanly (the
+one that does not, SM101A, needs its copybooks expanded first).
+
+[![Python 3.11+](https://img.shields.io/badge/python-3.11%20%7C%203.12%20%7C%203.13-blue.svg)](https://www.python.org/downloads/)
+[![PyPI](https://img.shields.io/pypi/v/cobol-parser.svg)](https://pypi.org/project/cobol-parser/)
+[![Coverage](https://img.shields.io/badge/coverage-90%25%20enforced-green.svg)](.okf/tests.md)
+[![License: AGPL v3](https://img.shields.io/badge/License-AGPL_v3-blue.svg)](LICENSE)
+
+## Install
+
+```bash
+pip install cobol-parser                # the library and the cobol-parser CLI
+pip install 'cobol-parser[mcp]'         # plus the MCP server
+pip install 'cobol-parser[jcl]'         # plus JCL (through jcl-parser)
+```
+
+Requires Python 3.11, 3.12 or 3.13. The core library has no runtime
+dependencies. Optional extras:
+
+| Extra | Adds |
+|---|---|
+| `mcp` | `cobol-parser mcp`, the MCP server ([docs/mcp-server.md](docs/mcp-server.md)) |
+| `jcl` | `cobol-parser jcl`, and the JCL joins in `field-impact`, `widen-check` and the MCP server |
+| `pl1`, `hlasm` | PL/I and assembler programs in the MCP server's call graph |
+| `fast` | Cython, to build the optional accelerators from source |
+| `dev`, `docs` | the test, lint and docs toolchain for working on this repo |
+
+To get the CLI on your `PATH` without a project virtualenv:
+
+```bash
+uv tool install cobol-parser
+uv tool install --editable .   # from a clone, picking up code changes
+```
+
+The install is pure Python by default. The PyPI wheels also carry three
+compiled Cython accelerators (about 3x lexer throughput), used only when
+`COBOLPARSER_ACCEL=1` is set at run time. From a clone, the same variable at
+build time compiles them:
+
+```bash
+pip install -e ".[fast]"
+COBOLPARSER_ACCEL=1 python setup.py build_ext --inplace
+export COBOLPARSER_ACCEL=1
+```
+
+## Quick start
+
+```python
+from cobolparser import parse_cobol
+
+result = parse_cobol(source, filename="prog.cbl")
+if result.success:
+    print(result.ast.program_name)
+    print(result.ast.to_json())
+for error in result.errors:
+    print(error)
+```
+
+`parse_cobol` stops at the first error by default. To collect every error and
+keep going, use `CobolParser(error_recovery=ErrorRecoveryMode.COLLECT_ALL)`.
+To expand `COPY` statements, use `parse_cobol_with_expansion(source,
+"prog.cbl", copybook_paths=["copybooks"])`.
+
+Every analyzer has the same shape, `<Analyzer>().analyze(ast)`, and returns a
+report with `to_dict()`:
+
+```python
+from cobolparser import LintAnalyzer, parse_cobol
+
+report = LintAnalyzer().analyze(parse_cobol(source, filename="prog.cbl").ast)
+for finding in report.findings:
+    print(finding.rule, finding.severity, finding.message)
+```
+
+## What it does
+
+The CLI has one subcommand per analyzer or generator. Every subcommand emits
+JSON by default, and takes `--help` with examples and its exit codes. The full
+list, with every flag, is in [.okf/cli.md](.okf/cli.md#subcommand-reference).
+
+### Check a program
+
+```bash
+cobol-parser lint prog.cbl --format text
+```
+
+```text
+prog.cbl:5:8: warning [unused-variables] WORKING-STORAGE item WS-NEVER-USED (level 01) is never referenced in the PROCEDURE DIVISION
+prog.cbl:10:12: error [move-type-check] Illegal MOVE: cannot move NUMERIC item 'WS-COUNTER' to ALPHABETIC item 'WS-ALPHA'
+prog.cbl:12:8: warning [dead-code] paragraph ORPHAN-PARA can never be reached
+3 finding(s): 1 error, 2 warning, 0 info (9 rule(s) ran)
+```
+
+`lint` runs nine analyzers over one parse: dead code, unused variables, MOVE
+type checks, I/O sequencing, LINKAGE checks, recursion risk, complexity,
+DCLGEN and SQL/CICS checks. `--rule` narrows the run, `--min-severity` the
+output, and `--fail-on error` exits 3 when a finding at that severity was
+reported, so it can gate a build. Each analyzer is also its own subcommand
+(`dead-code`, `xref`, `program-flow`, `complexity`, `health`, ...).
+
+### Check a whole codebase
+
+```bash
+cobol-parser scan --glob 'src/**/*.cbl' --analyzer health --clones --cache-dir .cache
+cobol-parser call-check src/ --copybook-path copybooks --fail-on error
+cobol-parser codebase-graph src/ --format mermaid
+```
+
+- `scan` runs analyzers over many files in one process, with per-file fault
+  isolation and cross-file duplicate-paragraph detection. With `--cache-dir`,
+  a file whose source, copybooks and copybook directories are unchanged reuses
+  its report from the last run.
+- `call-check` checks every `CALL` against the `PROCEDURE DIVISION USING` or
+  `ENTRY` it reaches in another file: argument count, passing mode
+  (`BY VALUE` / `BY REFERENCE`), byte length, USAGE and `RETURNING`. A
+  dynamic `CALL` is checked against every target its variable can hold.
+- `codebase-graph` and `copybook-impact` give the call graph and the programs
+  each copybook reaches.
+
+### Plan a change
+
+```bash
+cobol-parser lineage prog.cbl --field CUST-BAL
+cobol-parser field-impact CUST-ID src/ --jcl jcl/
+cobol-parser widen-check CUST-ID src/ --to '9(9)'
+cobol-parser diff old/prog.cbl new/prog.cbl
+```
+
+- `lineage` follows one field's values through a program, and `field-impact`
+  follows them across programs, files, DB2 columns and JCL datasets.
+- `widen-check` answers "what breaks if `CUST-ID` becomes `PIC 9(9)`": every
+  field downstream that would truncate the wider value, and, when the byte
+  length changes, the records whose later fields shift, the `REDEFINES` that
+  overlay it, the files whose LRECL changes (and the other programs reading
+  the same dataset), and the CALL / LINKAGE records on both sides.
+- `diff` compares two versions by structure, so reformatting and sequence
+  numbers are no change.
+
+### SARIF for code scanning
+
+`lint`, `call-check` and `scan` take `--format sarif` and write a SARIF 2.1.0
+log, which GitHub code scanning and IDE problem panes read directly:
+
+```bash
+cobol-parser lint prog.cbl --format sarif -o lint.sarif
+cobol-parser scan --glob 'src/**/*.cbl' --format sarif -o scan.sarif
+```
+
+### Generate and read test data
+
+```bash
+cobol-parser gen-data CUSTOMER.cpy --rows 100 --seed 7 -o customers.csv
+cobol-parser decode CUSTMAST.dat --copybook CUSTOMER.cpy --charset ebcdic
+cobol-parser compare expected.dat actual.dat --copybook CUSTOMER.cpy --key CUST-ID
+cobol-parser gen-fixtures prog.cbl --scenarios all -o bundle.json
+```
+
+`gen-data` writes synthetic records that fit each field's PICTURE, USAGE and
+level-88 values (a `seed` makes the output byte-identical across runs).
+`decode` and `compare` read record images, ASCII or EBCDIC, fixed or
+variable-length. `gen-fixtures` builds a test bundle for a whole program: one
+scenario per reachable branch, the values each mocked file, CALL, SQL or CICS
+boundary supplies, and the expectations to assert. Its schema is in
+[docs/fixture-bundle-schema.md](docs/fixture-bundle-schema.md).
+
+### Give an AI agent exact answers
+
+```bash
+pip install 'cobol-parser[mcp]'
+claude mcp add cobol -- cobol-parser mcp --root /path/to/checkout
+```
+
+`cobol-parser mcp` is a read-only Model Context Protocol server. Its tools
+answer "what is this program", "what is this field and where is it used",
+"who calls this", "what does a change to this copybook reach" and "show me
+this paragraph", with a `file:line` for every fact and a list of what they
+could not check. See [docs/mcp-server.md](docs/mcp-server.md).
+
+The repo also ships a `cobol-analysis` agent skill
+([.claude/skills/cobol-analysis/SKILL.md](.claude/skills/cobol-analysis/SKILL.md))
+that teaches an agent which subcommand answers which question.
+
+### Project config
+
+A `cobol.yml` at the project root sets copybook directories, source format,
+`>>DEFINE` symbols, data encodings and lint defaults once, for every
+subcommand. `cobol-parser config` shows which file is in effect. See
+[docs/project-config.md](docs/project-config.md).
+
+## Language coverage
+
+IBM Enterprise COBOL through 6.4: fixed and free format, `COPY ... REPLACING`
+and `REPLACE`, conditional compilation (`>>DEFINE`, `>>IF`, `>>EVALUATE`),
+nested and contained programs, object-oriented `CLASS-ID` / `METHOD-ID`,
+user-defined functions (`FUNCTION-ID`), DECLARATIVES, Report Writer, XML and
+JSON statements, and `EXEC SQL`, `EXEC CICS` and `EXEC DLI` blocks. One
+program per IBM program type (batch, VSAM, CICS, IMS, DB2, MQ, ...) runs
+through the whole stack in the test suite; the matrix is in
+[docs/enterprise-program-types.md](docs/enterprise-program-types.md).
+
+## Documentation
+
+| For | Read |
+|---|---|
+| Using the MCP server | [docs/mcp-server.md](docs/mcp-server.md) |
+| The `cobol.yml` project config | [docs/project-config.md](docs/project-config.md) |
+| The `gen-fixtures` bundle format | [docs/fixture-bundle-schema.md](docs/fixture-bundle-schema.md) |
+| Every subcommand and flag | [.okf/cli.md](.okf/cli.md) |
+| How the library works, area by area | [.okf/index.md](.okf/index.md) |
+| What the backend and mocky-mock rely on | [interface-contract.md](interface-contract.md) |
+| What changed in each release | [CHANGELOG.md](CHANGELOG.md) |
+| Everything under `docs/`, including design plans | [docs/README.md](docs/README.md) |
+
+The `.okf/` folder is the reference documentation: one topic file per area
+(lexer, parser, analyzers, generators, CLI, MCP, tests), written so that a
+person or an AI agent loads only the file for the task at hand.
+`scripts/okf_check.py` checks what those files claim against the code.
+
+## Development
+
+```bash
+pip install -e ".[dev,docs]"
+git config core.hooksPath .githooks              # the OKF pre-commit check
+
+python -m pytest tests/ -q -n 4 --no-cov         # fast full run
+python -m pytest tests/                          # with the 90% coverage floor
+COBOLPARSER_PERF=1 python -m pytest tests/performance/test_large_program.py   # 40k-line perf lane
+
+python -m ruff check .
+python -m black .
+python -m mypy cobolparser/
+```
+
+CI runs only on manual dispatch (`.github/workflows/ci.yml` has no push or PR
+trigger), so run lint, types and the full suite locally before pushing. Measure coverage on
+the pure-Python build: with `COBOLPARSER_ACCEL=1`, coverage.py cannot see the
+code the compiled modules replace. Test layout and markers:
+[.okf/tests.md](.okf/tests.md).
+
+## License
+
+GNU Affero General Public License v3.0 or later. Copyright © 2026 Samuel Dion.
+See [LICENSE](LICENSE).

cobol_parser-3.3.0/README.md

@@ -0,0 +1,251 @@
+# cobol-parser
+
+A COBOL parser for Python with an IBM Enterprise COBOL focus. It turns COBOL
+source into a traversable AST (JSON export, error recovery, copybook
+expansion with a source map back to the original lines), and builds static
+analyzers, test-data generators, a z/OS JCL front end and an MCP server for AI
+agents on top of it.
+
+458 of the 459 NIST COBOL-85 programs in `examples/cobol/` parse cleanly (the
+one that does not, SM101A, needs its copybooks expanded first).
+
+[![Python 3.11+](https://img.shields.io/badge/python-3.11%20%7C%203.12%20%7C%203.13-blue.svg)](https://www.python.org/downloads/)
+[![PyPI](https://img.shields.io/pypi/v/cobol-parser.svg)](https://pypi.org/project/cobol-parser/)
+[![Coverage](https://img.shields.io/badge/coverage-90%25%20enforced-green.svg)](.okf/tests.md)
+[![License: AGPL v3](https://img.shields.io/badge/License-AGPL_v3-blue.svg)](LICENSE)
+
+## Install
+
+```bash
+pip install cobol-parser                # the library and the cobol-parser CLI
+pip install 'cobol-parser[mcp]'         # plus the MCP server
+pip install 'cobol-parser[jcl]'         # plus JCL (through jcl-parser)
+```
+
+Requires Python 3.11, 3.12 or 3.13. The core library has no runtime
+dependencies. Optional extras:
+
+| Extra | Adds |
+|---|---|
+| `mcp` | `cobol-parser mcp`, the MCP server ([docs/mcp-server.md](docs/mcp-server.md)) |
+| `jcl` | `cobol-parser jcl`, and the JCL joins in `field-impact`, `widen-check` and the MCP server |
+| `pl1`, `hlasm` | PL/I and assembler programs in the MCP server's call graph |
+| `fast` | Cython, to build the optional accelerators from source |
+| `dev`, `docs` | the test, lint and docs toolchain for working on this repo |
+
+To get the CLI on your `PATH` without a project virtualenv:
+
+```bash
+uv tool install cobol-parser
+uv tool install --editable .   # from a clone, picking up code changes
+```
+
+The install is pure Python by default. The PyPI wheels also carry three
+compiled Cython accelerators (about 3x lexer throughput), used only when
+`COBOLPARSER_ACCEL=1` is set at run time. From a clone, the same variable at
+build time compiles them:
+
+```bash
+pip install -e ".[fast]"
+COBOLPARSER_ACCEL=1 python setup.py build_ext --inplace
+export COBOLPARSER_ACCEL=1
+```
+
+## Quick start
+
+```python
+from cobolparser import parse_cobol
+
+result = parse_cobol(source, filename="prog.cbl")
+if result.success:
+    print(result.ast.program_name)
+    print(result.ast.to_json())
+for error in result.errors:
+    print(error)
+```
+
+`parse_cobol` stops at the first error by default. To collect every error and
+keep going, use `CobolParser(error_recovery=ErrorRecoveryMode.COLLECT_ALL)`.
+To expand `COPY` statements, use `parse_cobol_with_expansion(source,
+"prog.cbl", copybook_paths=["copybooks"])`.
+
+Every analyzer has the same shape, `<Analyzer>().analyze(ast)`, and returns a
+report with `to_dict()`:
+
+```python
+from cobolparser import LintAnalyzer, parse_cobol
+
+report = LintAnalyzer().analyze(parse_cobol(source, filename="prog.cbl").ast)
+for finding in report.findings:
+    print(finding.rule, finding.severity, finding.message)
+```
+
+## What it does
+
+The CLI has one subcommand per analyzer or generator. Every subcommand emits
+JSON by default, and takes `--help` with examples and its exit codes. The full
+list, with every flag, is in [.okf/cli.md](.okf/cli.md#subcommand-reference).
+
+### Check a program
+
+```bash
+cobol-parser lint prog.cbl --format text
+```
+
+```text
+prog.cbl:5:8: warning [unused-variables] WORKING-STORAGE item WS-NEVER-USED (level 01) is never referenced in the PROCEDURE DIVISION
+prog.cbl:10:12: error [move-type-check] Illegal MOVE: cannot move NUMERIC item 'WS-COUNTER' to ALPHABETIC item 'WS-ALPHA'
+prog.cbl:12:8: warning [dead-code] paragraph ORPHAN-PARA can never be reached
+3 finding(s): 1 error, 2 warning, 0 info (9 rule(s) ran)
+```
+
+`lint` runs nine analyzers over one parse: dead code, unused variables, MOVE
+type checks, I/O sequencing, LINKAGE checks, recursion risk, complexity,
+DCLGEN and SQL/CICS checks. `--rule` narrows the run, `--min-severity` the
+output, and `--fail-on error` exits 3 when a finding at that severity was
+reported, so it can gate a build. Each analyzer is also its own subcommand
+(`dead-code`, `xref`, `program-flow`, `complexity`, `health`, ...).
+
+### Check a whole codebase
+
+```bash
+cobol-parser scan --glob 'src/**/*.cbl' --analyzer health --clones --cache-dir .cache
+cobol-parser call-check src/ --copybook-path copybooks --fail-on error
+cobol-parser codebase-graph src/ --format mermaid
+```
+
+- `scan` runs analyzers over many files in one process, with per-file fault
+  isolation and cross-file duplicate-paragraph detection. With `--cache-dir`,
+  a file whose source, copybooks and copybook directories are unchanged reuses
+  its report from the last run.
+- `call-check` checks every `CALL` against the `PROCEDURE DIVISION USING` or
+  `ENTRY` it reaches in another file: argument count, passing mode
+  (`BY VALUE` / `BY REFERENCE`), byte length, USAGE and `RETURNING`. A
+  dynamic `CALL` is checked against every target its variable can hold.
+- `codebase-graph` and `copybook-impact` give the call graph and the programs
+  each copybook reaches.
+
+### Plan a change
+
+```bash
+cobol-parser lineage prog.cbl --field CUST-BAL
+cobol-parser field-impact CUST-ID src/ --jcl jcl/
+cobol-parser widen-check CUST-ID src/ --to '9(9)'
+cobol-parser diff old/prog.cbl new/prog.cbl
+```
+
+- `lineage` follows one field's values through a program, and `field-impact`
+  follows them across programs, files, DB2 columns and JCL datasets.
+- `widen-check` answers "what breaks if `CUST-ID` becomes `PIC 9(9)`": every
+  field downstream that would truncate the wider value, and, when the byte
+  length changes, the records whose later fields shift, the `REDEFINES` that
+  overlay it, the files whose LRECL changes (and the other programs reading
+  the same dataset), and the CALL / LINKAGE records on both sides.
+- `diff` compares two versions by structure, so reformatting and sequence
+  numbers are no change.
+
+### SARIF for code scanning
+
+`lint`, `call-check` and `scan` take `--format sarif` and write a SARIF 2.1.0
+log, which GitHub code scanning and IDE problem panes read directly:
+
+```bash
+cobol-parser lint prog.cbl --format sarif -o lint.sarif
+cobol-parser scan --glob 'src/**/*.cbl' --format sarif -o scan.sarif
+```
+
+### Generate and read test data
+
+```bash
+cobol-parser gen-data CUSTOMER.cpy --rows 100 --seed 7 -o customers.csv
+cobol-parser decode CUSTMAST.dat --copybook CUSTOMER.cpy --charset ebcdic
+cobol-parser compare expected.dat actual.dat --copybook CUSTOMER.cpy --key CUST-ID
+cobol-parser gen-fixtures prog.cbl --scenarios all -o bundle.json
+```
+
+`gen-data` writes synthetic records that fit each field's PICTURE, USAGE and
+level-88 values (a `seed` makes the output byte-identical across runs).
+`decode` and `compare` read record images, ASCII or EBCDIC, fixed or
+variable-length. `gen-fixtures` builds a test bundle for a whole program: one
+scenario per reachable branch, the values each mocked file, CALL, SQL or CICS
+boundary supplies, and the expectations to assert. Its schema is in
+[docs/fixture-bundle-schema.md](docs/fixture-bundle-schema.md).
+
+### Give an AI agent exact answers
+
+```bash
+pip install 'cobol-parser[mcp]'
+claude mcp add cobol -- cobol-parser mcp --root /path/to/checkout
+```
+
+`cobol-parser mcp` is a read-only Model Context Protocol server. Its tools
+answer "what is this program", "what is this field and where is it used",
+"who calls this", "what does a change to this copybook reach" and "show me
+this paragraph", with a `file:line` for every fact and a list of what they
+could not check. See [docs/mcp-server.md](docs/mcp-server.md).
+
+The repo also ships a `cobol-analysis` agent skill
+([.claude/skills/cobol-analysis/SKILL.md](.claude/skills/cobol-analysis/SKILL.md))
+that teaches an agent which subcommand answers which question.
+
+### Project config
+
+A `cobol.yml` at the project root sets copybook directories, source format,
+`>>DEFINE` symbols, data encodings and lint defaults once, for every
+subcommand. `cobol-parser config` shows which file is in effect. See
+[docs/project-config.md](docs/project-config.md).
+
+## Language coverage
+
+IBM Enterprise COBOL through 6.4: fixed and free format, `COPY ... REPLACING`
+and `REPLACE`, conditional compilation (`>>DEFINE`, `>>IF`, `>>EVALUATE`),
+nested and contained programs, object-oriented `CLASS-ID` / `METHOD-ID`,
+user-defined functions (`FUNCTION-ID`), DECLARATIVES, Report Writer, XML and
+JSON statements, and `EXEC SQL`, `EXEC CICS` and `EXEC DLI` blocks. One
+program per IBM program type (batch, VSAM, CICS, IMS, DB2, MQ, ...) runs
+through the whole stack in the test suite; the matrix is in
+[docs/enterprise-program-types.md](docs/enterprise-program-types.md).
+
+## Documentation
+
+| For | Read |
+|---|---|
+| Using the MCP server | [docs/mcp-server.md](docs/mcp-server.md) |
+| The `cobol.yml` project config | [docs/project-config.md](docs/project-config.md) |
+| The `gen-fixtures` bundle format | [docs/fixture-bundle-schema.md](docs/fixture-bundle-schema.md) |
+| Every subcommand and flag | [.okf/cli.md](.okf/cli.md) |
+| How the library works, area by area | [.okf/index.md](.okf/index.md) |
+| What the backend and mocky-mock rely on | [interface-contract.md](interface-contract.md) |
+| What changed in each release | [CHANGELOG.md](CHANGELOG.md) |
+| Everything under `docs/`, including design plans | [docs/README.md](docs/README.md) |
+
+The `.okf/` folder is the reference documentation: one topic file per area
+(lexer, parser, analyzers, generators, CLI, MCP, tests), written so that a
+person or an AI agent loads only the file for the task at hand.
+`scripts/okf_check.py` checks what those files claim against the code.
+
+## Development
+
+```bash
+pip install -e ".[dev,docs]"
+git config core.hooksPath .githooks              # the OKF pre-commit check
+
+python -m pytest tests/ -q -n 4 --no-cov         # fast full run
+python -m pytest tests/                          # with the 90% coverage floor
+COBOLPARSER_PERF=1 python -m pytest tests/performance/test_large_program.py   # 40k-line perf lane
+
+python -m ruff check .
+python -m black .
+python -m mypy cobolparser/
+```
+
+CI runs only on manual dispatch (`.github/workflows/ci.yml` has no push or PR
+trigger), so run lint, types and the full suite locally before pushing. Measure coverage on
+the pure-Python build: with `COBOLPARSER_ACCEL=1`, coverage.py cannot see the
+code the compiled modules replace. Test layout and markers:
+[.okf/tests.md](.okf/tests.md).
+
+## License
+
+GNU Affero General Public License v3.0 or later. Copyright © 2026 Samuel Dion.
+See [LICENSE](LICENSE).