feat: referentiel projets dynamique + validation nommage by design

Durcit la convention de nommage des projets (dérive constatée : 'Sliding Automation', 'code_versioning'... au lieu des formes canoniques). - trilium_api.py : projets_canoniques() lit le référentiel = valeurs du label projet sur les notes de type=projet (source unique, pas de constante en dur). Note-projet CodeVersioning créée (manquait). - mcp_server.py : _valider_projet() branché dans les 6 tools de création (add_decision/history/backlog, new_conversation, create_entite, add_skill). Refuse un projet non canonique (suggestion si faute) ou inconnu (renvoi au processus de création de projet). Ne verrouille pas si référentiel illisible. - lint_audit.py : VAL-nommage aligné sur le référentiel (attrape casse, espace ET snake_case ; l'ancien 'contient un espace' ratait code_versioning). - Données : 79 notes ré-étiquetées vers les 3 formes canoniques. Quality by design : l'erreur de nommage devient impossible à l'écriture, le Lint n'est plus que le filet de sécurité.
2026-07-17 14:59:33 +02:00
parent d1e67431b6
commit 943acbc573
2425 changed files with 710525 additions and 6 deletions
@@ -0,0 +1,118 @@
+#cython: language_level=3
+
+from __future__ import print_function
+from cpython.mem cimport PyMem_Malloc, PyMem_Free
+from cpython cimport PyObject_GetBuffer, PyBuffer_Release, PyBUF_SIMPLE, \
+                     Py_buffer
+
+from .errors import HttpParserInvalidURLError
+
+cimport cython
+from . cimport url_cparser as uparser
+
+__all__ = ('parse_url',)
+
+DEF MAX_URL_LENGTH = (1 << 16) - 1
+
+@cython.freelist(250)
+cdef class URL:
+    cdef readonly bytes schema
+    cdef readonly bytes host
+    cdef readonly object port
+    cdef readonly bytes path
+    cdef readonly bytes query
+    cdef readonly bytes fragment
+    cdef readonly bytes userinfo
+
+    def __cinit__(self, bytes schema, bytes host, object port, bytes path,
+                  bytes query, bytes fragment, bytes userinfo):
+
+        self.schema = schema
+        self.host = host
+        self.port = port
+        self.path = path
+        self.query = query
+        self.fragment = fragment
+        self.userinfo = userinfo
+
+    def __repr__(self):
+        return ('<URL schema: {!r}, host: {!r}, port: {!r}, path: {!r}, '
+                'query: {!r}, fragment: {!r}, userinfo: {!r}>'
+                .format(self.schema, self.host, self.port, self.path,
+                    self.query, self.fragment, self.userinfo))
+
+
+def parse_url(url):
+    cdef:
+        Py_buffer py_buf
+        char* buf_data
+        uparser.http_parser_url* parsed
+        int res
+        bytes schema = None
+        bytes host = None
+        object port = None
+        bytes path = None
+        bytes query = None
+        bytes fragment = None
+        bytes userinfo = None
+        object result = None
+        int off
+        int ln
+
+    parsed = <uparser.http_parser_url*> \
+                        PyMem_Malloc(sizeof(uparser.http_parser_url))
+    uparser.http_parser_url_init(parsed)
+
+    PyObject_GetBuffer(url, &py_buf, PyBUF_SIMPLE)
+    try:
+        if py_buf.len > MAX_URL_LENGTH:
+            # http_parser stores URL field offsets/lengths as uint16_t,
+            # so URLs longer than this will cause silent truncation.
+            # See https://github.com/MagicStack/httptools/issues/142
+            raise HttpParserInvalidURLError(
+                "url is too long: url length of {} bytes exceeds the "
+                "maximum of {} bytes".format(py_buf.len, MAX_URL_LENGTH))
+
+        buf_data = <char*>py_buf.buf
+        res = uparser.http_parser_parse_url(buf_data, py_buf.len, 0, parsed)
+
+        if res == 0:
+            if parsed.field_set & (1 << uparser.UF_SCHEMA):
+                off = parsed.field_data[<int>uparser.UF_SCHEMA].off
+                ln = parsed.field_data[<int>uparser.UF_SCHEMA].len
+                schema = buf_data[off:off+ln]
+
+            if parsed.field_set & (1 << uparser.UF_HOST):
+                off = parsed.field_data[<int>uparser.UF_HOST].off
+                ln = parsed.field_data[<int>uparser.UF_HOST].len
+                host = buf_data[off:off+ln]
+
+            if parsed.field_set & (1 << uparser.UF_PORT):
+                port = parsed.port
+
+            if parsed.field_set & (1 << uparser.UF_PATH):
+                off = parsed.field_data[<int>uparser.UF_PATH].off
+                ln = parsed.field_data[<int>uparser.UF_PATH].len
+                path = buf_data[off:off+ln]
+
+            if parsed.field_set & (1 << uparser.UF_QUERY):
+                off = parsed.field_data[<int>uparser.UF_QUERY].off
+                ln = parsed.field_data[<int>uparser.UF_QUERY].len
+                query = buf_data[off:off+ln]
+
+            if parsed.field_set & (1 << uparser.UF_FRAGMENT):
+                off = parsed.field_data[<int>uparser.UF_FRAGMENT].off
+                ln = parsed.field_data[<int>uparser.UF_FRAGMENT].len
+                fragment = buf_data[off:off+ln]
+
+            if parsed.field_set & (1 << uparser.UF_USERINFO):
+                off = parsed.field_data[<int>uparser.UF_USERINFO].off
+                ln = parsed.field_data[<int>uparser.UF_USERINFO].len
+                userinfo = buf_data[off:off+ln]
+
+            return URL(schema, host, port, path, query, fragment, userinfo)
+        else:
+            raise HttpParserInvalidURLError("invalid url {!r}".format(url))
+    finally:
+        PyBuffer_Release(&py_buf)
+        PyMem_Free(parsed)