mcaxl/tests/test_sql_validator.py
Ryan Malloy ca6956e826 Rename to mcaxl + scrub for public PyPI release
Renames the package from `mcp-cucm-axl` to `mcaxl` to fit the
operator's mc<interface> naming convention (mcusb, mcaxl, …),
and scrubs Bingham-specific defaults so the package works for
anyone, anywhere.

Rename:
  - pyproject.toml: name, scripts entry point, description
  - src/mcp_cucm_axl/ → src/mcaxl/ (git mv preserves history)
  - All Python imports updated via sed
  - Cache directory: ~/.cache/mcp-cucm-axl/ → ~/.cache/mcaxl/
  - Log prefix [mcp-cucm-axl] → [mcaxl]
  - Package version lookup: importlib.metadata.version("mcaxl")
  - .mcp.json command updated to invoke `mcaxl` script
  - All 155 tests pass under the new name (verified)

Bingham-specific scrubs:
  - docs_loader._DEFAULT_INDEX_DIR: hardcoded /home/rpm/bingham/...
    path removed; defaults to None. Operators set CISCO_DOCS_INDEX_PATH
    env var; without it, prompts gracefully degrade with a fallback
    notice instructing the LLM to use the cisco-docs MCP search_docs
    tool instead.
  - prompts/_common.docs_or_empty_msg: removed the explicit
    /home/rpm/bingham/... path from the fallback message text.
  - server.py: removed dead-code copy of _docs_or_empty_msg() that
    was leftover from before the prompts package extraction.
  - README.md: completely rewritten as a public-facing readme. Lead
    paragraph names CUCM as the target platform, install instructions
    cover uvx / pip / Claude Code MCP add. Recommends cisco-cucm-mcp
    as the operations counterpart.

PyPI metadata:
  - Initial CalVer version: 2026.04.27
  - License: MIT (LICENSE file added)
  - Project URLs: Homepage / Source / Issues / Changelog all point
    at git.supported.systems/mcp/mcaxl (newly-created Gitea repo
    in the mcp/ org for PyPI releases)
  - Classifiers: Beta / Telecommunications Industry / Topic:Telephony
  - Keywords: mcp, cisco, cucm, axl, risport, voip, sip, audit
  - sdist excludes: CLAUDE.md, .env*, axlsqltoolkit.zip, audits/,
    tests/, pytest/ruff caches. Verified clean: wheel ships only the
    mcaxl/ source tree + LICENSE + METADATA + entry_points.

CHANGELOG.md added with a 2026.04.27 initial-release entry,
documenting tool/prompt counts, structural read-only guarantees,
Hamilton review closure, live-cluster verification, and known
limitations.

Build verification:
  - `uv build` produces clean wheel + sdist
  - Wheel: 22 source files, 195KB total, no Bingham-specific files
  - Sdist excludes verified: no CLAUDE.md, no axlsqltoolkit.zip
  - Entry point: `mcaxl = mcaxl.server:main`
  - Package installs as mcaxl==2026.4.27
2026-04-27 12:53:54 -06:00

178 lines
7.4 KiB
Python

"""Tests for the SELECT-only SQL guardrail."""
import pytest
from mcaxl.sql_validator import validate_select, SqlValidationError
class TestSelectAccepted:
def test_simple_select(self):
assert validate_select("SELECT * FROM device") == "SELECT * FROM device"
def test_with_cte(self):
q = "WITH x AS (SELECT 1 FROM systables) SELECT * FROM x"
assert validate_select(q) == q
def test_lowercase_select(self):
assert validate_select("select * from numplan") == "select * from numplan"
def test_trailing_semicolon_stripped(self):
assert validate_select("SELECT 1 FROM device;") == "SELECT 1 FROM device"
def test_block_comments_stripped(self):
q = "/* comment */ SELECT 1 FROM device"
cleaned = validate_select(q)
assert "SELECT 1 FROM device" in cleaned
def test_line_comments_stripped(self):
q = "-- a comment\nSELECT 1 FROM device"
cleaned = validate_select(q)
assert "SELECT 1 FROM device" in cleaned
class TestRejected:
def test_empty(self):
with pytest.raises(SqlValidationError, match="empty"):
validate_select("")
def test_whitespace_only(self):
with pytest.raises(SqlValidationError, match="empty"):
validate_select(" \n ")
def test_only_comments(self):
with pytest.raises(SqlValidationError, match="empty"):
validate_select("-- just a comment\n/* and another */")
def test_insert_rejected(self):
with pytest.raises(SqlValidationError, match="INSERT"):
validate_select("INSERT INTO device VALUES (1)")
def test_update_rejected(self):
with pytest.raises(SqlValidationError, match="UPDATE"):
validate_select("UPDATE device SET name='x' WHERE pkid='y'")
def test_delete_rejected(self):
with pytest.raises(SqlValidationError, match="DELETE"):
validate_select("DELETE FROM device WHERE pkid='y'")
def test_drop_rejected(self):
with pytest.raises(SqlValidationError, match="DROP"):
validate_select("DROP TABLE device")
def test_select_with_embedded_drop_rejected(self):
# Belt-and-suspenders: even if "DROP" appears in a quoted string-ish
# position our keyword filter still catches it. AXL would also reject
# this, but failing fast on the client saves a SOAP round-trip.
with pytest.raises(SqlValidationError, match="DROP"):
validate_select("SELECT 1 FROM device; DROP TABLE device")
def test_truncate_rejected(self):
with pytest.raises(SqlValidationError, match="TRUNCATE"):
validate_select("TRUNCATE TABLE device")
class TestEdgeCases:
def test_keyword_as_column_name_blocked(self):
# A column named "delete" would be blocked. This is acceptable —
# the data dictionary doesn't use SQL keywords as column names,
# and conservative blocking is the right call for v1.
with pytest.raises(SqlValidationError):
validate_select("SELECT delete FROM device")
def test_select_with_subquery(self):
q = "SELECT name FROM device WHERE pkid IN (SELECT fkdevice FROM numplan)"
assert "SELECT name FROM device" in validate_select(q)
class TestStringLiterals:
"""Forbidden keywords inside string literals must be ignored.
Otherwise CSS names like 'Call Forward-CSS', DN descriptions containing
'DELETE' (e.g., 'Delete this voicemail line'), or partition names with
'INSERT' would all fail to query, even though the SQL itself is read-only.
"""
def test_call_inside_string_literal_passes(self):
q = "SELECT pkid FROM callingsearchspace WHERE name = 'Call Forward-CSS'"
result = validate_select(q)
assert "Call Forward-CSS" in result # literal preserved
def test_delete_inside_string_literal_passes(self):
q = "SELECT pkid FROM numplan WHERE description = 'Delete after audit'"
result = validate_select(q)
assert "Delete after audit" in result
def test_drop_inside_string_literal_passes(self):
q = "SELECT pkid FROM numplan WHERE description = 'DROP table backup'"
assert validate_select(q)
def test_actual_drop_outside_literal_still_blocked(self):
with pytest.raises(SqlValidationError, match="DROP"):
validate_select("SELECT 1 FROM device; DROP TABLE backup")
def test_escaped_quote_in_literal(self):
# Informix uses '' (doubled) as escaped single quote within literals
q = "SELECT pkid FROM numplan WHERE description = 'O''Brien''s line'"
result = validate_select(q)
assert "O''Brien''s line" in result
def test_keyword_just_outside_literal_blocked(self):
# The literal 'safe text' is fine; the trailing DROP is not.
with pytest.raises(SqlValidationError, match="DROP"):
validate_select("SELECT 1 FROM device WHERE x = 'safe text' OR DROP")
def test_multiple_literals(self):
q = "SELECT 1 FROM numplan WHERE name = 'CALL' AND description = 'UPDATE pending'"
assert validate_select(q)
class TestLiteralPreservedInOutput:
"""Hamilton review CRITICAL #1: comment-strip mutated string literals.
The query SENT to AXL must preserve the literal contents byte-for-byte.
Previously, the comment-strip pass ran before the literal-aware pass,
so `--` or `/* */` inside a quoted string were silently eaten on the
way to the cluster. An LLM dialing `description LIKE '%-- old%'` got
a different query than it asked for.
"""
def test_dash_dash_inside_literal_preserved(self):
q = "SELECT * FROM numplan WHERE description = 'Smith -- old line'"
result = validate_select(q)
assert "Smith -- old line" in result, (
f"line-comment marker inside literal must NOT be stripped; got: {result!r}"
)
def test_block_comment_marker_inside_literal_preserved(self):
q = "SELECT * FROM device WHERE name = 'before /* still in literal */ after'"
result = validate_select(q)
assert "/* still in literal */" in result
assert "before" in result and "after" in result
def test_like_pattern_with_dash_dash_preserved(self):
# Real-world case: an LLM searches for descriptions containing "--"
q = "SELECT pkid FROM numplan WHERE description LIKE '%-- old%'"
result = validate_select(q)
assert "'%-- old%'" in result
def test_actual_line_comment_outside_literal_still_handled(self):
# An actual --comment outside any literal is fine (AXL handles it),
# and the keyword check ignores it.
q = "SELECT 1 FROM device -- a real comment at the end"
result = validate_select(q)
# We don't strip from output, so the comment stays in the returned text.
# The important thing is the validator passes and a forbidden keyword
# in the comment wouldn't trip the check (covered separately).
assert "SELECT 1 FROM device" in result
def test_forbidden_keyword_inside_real_comment_does_not_trip(self):
# Real comment, with a forbidden keyword in it, should not trip the validator
q = "SELECT 1 FROM device -- TODO: someone DELETE the old test data"
result = validate_select(q)
assert "SELECT 1" in result
def test_block_literal_with_drop_inside_preserved(self):
q = "SELECT 1 FROM numplan WHERE description = 'log: DROP detected'"
result = validate_select(q)
assert "'log: DROP detected'" in result