DEV-1520: add SQL Server (T-SQL) datasource support#162
Merged
ZmeiGorynych merged 12 commits intoJun 3, 2026
Conversation
- SQL generation: DATETRUNC/DATEADD for date ops, STDEV/STDEVP/VAR/VARP for 1-arg stat aggs, variance-decomposition formula for corr/covar_samp/ covar_pop (same formula now used for MySQL too); NotImplementedError for median/percentile (PERCENTILE_CONT is window-only in T-SQL) - DatasourceConfig: mssql/sqlserver/tsql types generate mssql+pyodbc:// connection strings with ODBC Driver 18 and TrustServerCertificate=yes - Ingestion: _SA_TYPE_MAP gains T-SQL-specific SA types (TINYINT, DATETIME2, SMALLDATETIME, DATETIMEOFFSET, NVARCHAR, NCHAR, NTEXT, MONEY, SMALLMONEY) - pyproject.toml: pyodbc optional dep + sqlserver extra - examples/sqlserver/: Docker Compose example with custom Dockerfile that installs msodbcsql18, SQL Server 2022 container, seed + verify scripts - examples/seed.py: CREATE_SQL_TSQL branch (NVARCHAR, DATETIME2) - examples/mysql/verify.py: now calls check_corr_covar() (MySQL implements it) - 65 new tests covering all T-SQL generation paths and type mappings Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Contributor
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (10)
📝 WalkthroughWalkthroughAdds full SQL Server (T‑SQL) support: promote SQL Server to Tier 1, add mssql+pyodbc connection strings, extend ingestion type mappings, implement T‑SQL date/time and statistical SQL (variance‑decomposition for corr/covar), add a SQL Server Docker example and scripts, and expand tests. ChangesSQL Server Full Support Implementation
Sequence Diagram(s) sequenceDiagram
participant examples_verify as examples/sqlserver/verify.py
participant verify_common as examples/verify_common.py
participant sql_generator as slayer/sql/generator.py
participant sqlserver as SQLServer
examples_verify->>verify_common: run common checks & corr/covar
verify_common->>sql_generator: request corr/covar SQL (variance-decomposition)
sql_generator->>sqlserver: emit T-SQL query (DATEADD/DATETRUNC or formula SQL)
🎯 4 (Complex) | ⏱️ ~50 minutes Possibly Related PRs
Suggested Reviewers
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 5
🧹 Nitpick comments (1)
tests/test_models.py (1)
838-846: ⚡ Quick winMake ODBC driver format assertion more precise.
The test checks for "odbc", "driver", and "18" separately, but doesn't verify the exact format required by the coding guideline:
ODBC+Driver+18+for+SQL+Server. A more precise assertion would prevent future regressions where someone might break the driver string format.✨ Suggested improvement
def test_sqlserver_connection_string_includes_odbc_driver_param(self) -> None: ds = DatasourceConfig(name="test", type="mssql", host="sqlhost", database="db") cs = ds.get_connection_string() - # Must include exact ODBC driver name so pyodbc can locate the installed driver - cs_lower = cs.lower() - assert "odbc" in cs_lower - assert "driver" in cs_lower - assert "18" in cs # msodbcsql18 is the expected driver + # Must include exact ODBC driver name as specified in coding guidelines + assert "driver=ODBC+Driver+18+for+SQL+Server" in csAs per coding guidelines:
examples/sqlserver/**/*.py: Pin ODBC driver version to 18 in pyodbc connection strings; use formatODBC+Driver+18+for+SQL+Server.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_models.py` around lines 838 - 846, Update the test_sqlserver_connection_string_includes_odbc_driver_param to assert the exact ODBC driver substring expected by our guideline instead of separate checks; call DatasourceConfig(name="test", type="mssql", host="sqlhost", database="db").get_connection_string() as before but replace the three separate asserts with a single assertion that the connection string contains the exact token "ODBC+Driver+18+for+SQL+Server" (case-insensitive check is fine), so the test will fail if the driver format deviates.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@examples/sqlserver/Dockerfile`:
- Line 23: The Dockerfile currently redundantly installs pyodbc via the pip
command that contains "pip install --no-cache-dir psycopg2-binary pymysql
clickhouse-sqlalchemy pyodbc && \"; remove pyodbc from that pip install line so
the image relies on the dependency brought in by "poetry install -E all" (which
already includes the sqlserver extra), and ensure the Dockerfile keeps the other
DB drivers (psycopg2-binary, pymysql, clickhouse-sqlalchemy) intact; run a build
to verify no missing-dependency errors and that pyodbc is resolved via poetry's
extras.
- Around line 19-24: The Docker build is failing because poetry.lock is out of
sync with pyproject.toml; regenerate and commit an updated lock file by running
poetry lock from the repository root (so the COPY step in Dockerfile will
include a fresh poetry.lock). After running `poetry lock` (no-update if you want
to preserve versions), verify the updated poetry.lock is committed to the repo
so the Dockerfile's COPY pyproject.toml poetry.lock* step uses the synchronized
lockfile.
In `@examples/sqlserver/start.sh`:
- Around line 7-17: The script mixes sync/async entrypoints: ingest_datasource
is a synchronous function in slayer/engine/ingestion.py so remove the
run_sync(...) wrapper and call ingest_datasource(...) directly (or
alternatively, if you intended to use the async higher-level API
ingest_datasource_models from slayer/mcp/server.py then import
ingest_datasource_models and wrap that call with run_sync(...) instead); update
the imports and the call site accordingly (look for ingest_datasource,
ingest_datasource_models, run_sync, DatasourceConfig, YAMLStorage) so the chosen
function's sync/async contract is used consistently.
In `@examples/sqlserver/verify.py`:
- Around line 28-47: The docstring for check_column_types is misleading about
expected type strings; update the docstring in examples/verify_common.py to
state that expected_types should use the canonical DataType enum .value strings
(e.g., "INT", "TEXT", "TIMESTAMP", "DOUBLE") as produced in
examples/sqlserver/verify.py (or explicitly document that legacy lowercase
aliases like "number"/"string"/"time"/"date" will be mapped to the canonical
values), and mention the related symbols check_column_types, expected_types and
DataType so callers know which exact string values are valid.
In `@pyproject.toml`:
- Line 48: The lockfile is out of sync with pyproject.toml (the pyodbc optional
dependency and other changes at the referenced lines), so regenerate and commit
an updated poetry.lock: run `poetry lock` locally (or CI equivalent) to rebuild
the lockfile, verify it updated entries for pyodbc and the other modified
dependencies, then add and commit the regenerated poetry.lock so the pipeline
uses the synced lockfile.
---
Nitpick comments:
In `@tests/test_models.py`:
- Around line 838-846: Update the
test_sqlserver_connection_string_includes_odbc_driver_param to assert the exact
ODBC driver substring expected by our guideline instead of separate checks; call
DatasourceConfig(name="test", type="mssql", host="sqlhost",
database="db").get_connection_string() as before but replace the three separate
asserts with a single assertion that the connection string contains the exact
token "ODBC+Driver+18+for+SQL+Server" (case-insensitive check is fine), so the
test will fail if the driver format deviates.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 9a91a39e-4b97-40bc-88f3-85d65fbb5700
📒 Files selected for processing (18)
CLAUDE.mdexamples/mysql/verify.pyexamples/seed.pyexamples/sqlserver/CLAUDE.mdexamples/sqlserver/Dockerfileexamples/sqlserver/README.mdexamples/sqlserver/docker-compose.ymlexamples/sqlserver/slayer_data/.gitignoreexamples/sqlserver/start.shexamples/sqlserver/verify.pyexamples/verify_common.pypyproject.tomlslayer/core/models.pyslayer/engine/ingestion.pyslayer/sql/generator.pytests/test_ingestion.pytests/test_models.pytests/test_sql_generator.py
…nerate poetry.lock poetry.lock was out of sync after adding pyodbc to pyproject.toml. Also removed pyodbc from the manual pip install layer (it's already pulled in by `poetry install -E all`) and merged the two RUN instructions that install + create the system user into one. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Group 2 — correctness: - generator.py _build_date_trunc (tsql): wrap non-Column/Cast expressions in CAST(... AS TIMESTAMP) before passing to DATETRUNC; plain function-call results aren't implicitly temporal in T-SQL. - generator.py _build_covar_formula: guard corr denominator with NULLIF(..., 0) so constant columns (STDEV=0) return NULL instead of a divide-by-zero error. - ingestion.py: add MONEY/SMALLMONEY to _FLOAT_LIKE_SA_TYPES so monetary columns get NumberFormat.FLOAT instead of NumberFormat.INT. Group 3 — Sonar / code quality: - models.py: extract _get_tsql_connection_string() helper; reduces cognitive complexity of get_connection_string from 18 to 14 (Sonar S3776). - tests/test_models.py: add NOSONAR(S2068) for test-only fixture credential; tighten ODBC driver assertion to the exact string. - examples/verify_common.py: update check_column_types docstring to show current uppercase DataType values (TEXT/DOUBLE/TIMESTAMP/DATE). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@slayer/core/models.py`:
- Around line 750-762: The _get_tsql_connection_string function manually
concatenates username, password, host, port and database into an mssql+pyodbc
URI which can break with reserved URL characters; replace the manual
interpolation by constructing the connection URL with SQLAlchemy's
URL.create(...) and call render_as_string() to ensure proper
quoting/encoding—use URL.create(drivername="mssql+pyodbc",
username=self.username, password=self.password, host=self.host or "localhost",
port=self.port, database=self.database or "", query={"driver":"ODBC Driver 18
for SQL Server", "TrustServerCertificate":"yes"}) in _get_tsql_connection_string
so SQLAlchemy handles encoding (and remove manual auth/params concatenation).
In `@slayer/sql/generator.py`:
- Around line 1830-1832: The DATETRUNC branch currently wraps
non-Column/non-Cast expressions by casting to TIMESTAMP (exp.Cast with
exp.DataType.build("TIMESTAMP")), but in T-SQL TIMESTAMP is rowversion
(non-temporal) and invalid for DATETRUNC; update the fallback cast to a temporal
SQL Server type such as DATETIME2 by replacing exp.DataType.build("TIMESTAMP")
with exp.DataType.build("DATETIME2") where col_expr is wrapped (the exp.Cast
creation around col_expr in the DATETRUNC handling).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 076bbc64-1ecf-40da-be1c-e31a0eff32cc
⛔ Files ignored due to path filters (1)
poetry.lockis excluded by!**/*.lock
📒 Files selected for processing (6)
examples/sqlserver/Dockerfileexamples/verify_common.pyslayer/core/models.pyslayer/engine/ingestion.pyslayer/sql/generator.pytests/test_models.py
✅ Files skipped from review due to trivial changes (1)
- examples/verify_common.py
🚧 Files skipped from review as they are similar to previous changes (3)
- examples/sqlserver/Dockerfile
- slayer/engine/ingestion.py
- tests/test_models.py
Group 1 — connection string safety:
- models.py: replace manual URL interpolation in _get_tsql_connection_string()
with sqlalchemy.engine.URL.create(); passwords containing reserved characters
like '@' (e.g. the Docker example's 'YourStrong@Passw0rd') are now properly
percent-encoded instead of breaking URL parsing.
- docker-compose.yml: URL-encode '@' in the seed container's literal connection
string ('%40Passw0rd') so SQLAlchemy parses host/password correctly.
- tests: add test_sqlserver_special_chars_in_password_are_url_encoded to pin
the fix.
Group 2 — BIT type mapping:
- ingestion.py: add "BIT": DataType.BOOLEAN to _SA_TYPE_MAP; SQL Server's BIT
type is reflected by SQLAlchemy as class name "BIT", but was missing from the
map and falling back to DataType.TEXT instead of DataType.BOOLEAN.
- tests: add test_tsql_bit_maps_to_boolean.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
examples/sqlserver/docker-compose.yml (2)
29-29:⚠️ Potential issue | 🟠 Major | ⚡ Quick winMake database creation idempotent for reruns.
Line 29 uses unconditional
CREATE DATABASE slayer_demo; rerunning the stack can fail atcreatedband block dependent services. Wrap with existence check.Suggested change
- -Q "CREATE DATABASE slayer_demo" -No + -Q "IF DB_ID(N'slayer_demo') IS NULL CREATE DATABASE slayer_demo;" -No🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/sqlserver/docker-compose.yml` at line 29, Replace the unconditional CREATE statement so database creation is idempotent: locate the sqlcmd invocation that contains the string "CREATE DATABASE slayer_demo" (the -Q argument) and change it to perform a conditional check before creating the DB (for example using a guard like IF DB_ID('slayer_demo') IS NULL CREATE DATABASE slayer_demo or an equivalent existence-check SQL) so repeated stack runs do not error when the database already exists.
6-6:⚠️ Potential issue | 🟠 Major | ⚡ Quick winRemove hardcoded database credentials from compose and command args.
Line 6 and Line 40 embed production-grade secrets directly in the file/CLI, which is easy to leak in git history, logs, and process lists. Move these to environment variables (or
.env) and reference them in commands/URLs.Suggested change
services: sqlserver: @@ environment: ACCEPT_EULA: "Y" - MSSQL_SA_PASSWORD: "YourStrong@Passw0rd" + MSSQL_SA_PASSWORD: "${MSSQL_SA_PASSWORD:?set in .env}" MSSQL_PID: "Developer" @@ createdb: @@ command: > /opt/mssql-tools18/bin/sqlcmd - -S sqlserver -U sa -P 'YourStrong@Passw0rd' + -S sqlserver -U sa -P "${MSSQL_SA_PASSWORD}" -Q "CREATE DATABASE slayer_demo" -No @@ seed: @@ command: > python /examples/seed.py - "mssql+pyodbc://sa:YourStrong%40Passw0rd@sqlserver:1433/slayer_demo?driver=ODBC+Driver+18+for+SQL+Server&TrustServerCertificate=yes" + "mssql+pyodbc://sa:${MSSQL_SA_PASSWORD_ENCODED}`@sqlserver`:1433/slayer_demo?driver=ODBC+Driver+18+for+SQL+Server&TrustServerCertificate=yes"Also applies to: 15-16, 28-29, 40-40
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/sqlserver/docker-compose.yml` at line 6, The compose file currently embeds production secrets (e.g., the MSSQL_SA_PASSWORD environment key and other DB credentials used in command/URL arguments) directly in docker-compose.yml; replace all hardcoded passwords with environment variable references (for example use a named env var like MSSQL_SA_PASSWORD or DB_PASSWORD loaded from a .env or host env) and update any command/URL arguments that contain credentials to reference those env vars instead (search for MSSQL_SA_PASSWORD and any DB URL/command args in the file and swap literals for ${VAR_NAME}); ensure the .env or deployment docs define the variables and remove plaintext secrets from the repo and CLI/compose files.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@examples/sqlserver/docker-compose.yml`:
- Line 29: Replace the unconditional CREATE statement so database creation is
idempotent: locate the sqlcmd invocation that contains the string "CREATE
DATABASE slayer_demo" (the -Q argument) and change it to perform a conditional
check before creating the DB (for example using a guard like IF
DB_ID('slayer_demo') IS NULL CREATE DATABASE slayer_demo or an equivalent
existence-check SQL) so repeated stack runs do not error when the database
already exists.
- Line 6: The compose file currently embeds production secrets (e.g., the
MSSQL_SA_PASSWORD environment key and other DB credentials used in command/URL
arguments) directly in docker-compose.yml; replace all hardcoded passwords with
environment variable references (for example use a named env var like
MSSQL_SA_PASSWORD or DB_PASSWORD loaded from a .env or host env) and update any
command/URL arguments that contain credentials to reference those env vars
instead (search for MSSQL_SA_PASSWORD and any DB URL/command args in the file
and swap literals for ${VAR_NAME}); ensure the .env or deployment docs define
the variables and remove plaintext secrets from the repo and CLI/compose files.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: d943f145-8503-4ee3-b937-c11115639e83
📒 Files selected for processing (5)
examples/sqlserver/docker-compose.ymlslayer/core/models.pyslayer/engine/ingestion.pytests/test_ingestion.pytests/test_models.py
✅ Files skipped from review due to trivial changes (1)
- tests/test_models.py
🚧 Files skipped from review as they are similar to previous changes (3)
- tests/test_ingestion.py
- slayer/engine/ingestion.py
- slayer/core/models.py
Finding 1 — docker-compose.yml:
- Wrap CREATE DATABASE in an existence guard so reruns don't fail:
IF DB_ID(N'slayer_demo') IS NULL CREATE DATABASE slayer_demo
Finding 2 — sql_client.py:
- T-SQL (SQL Server) does not support LIMIT; the type-probing query
SELECT * FROM (...) AS _types LIMIT 0 was a syntax error, causing
get_column_types to silently return {} for every SQL Server call.
- Extract _build_type_probe_sql() helper that emits
SELECT TOP N * FROM (...) AS _types for mssql/sqlserver/tsql and
the existing LIMIT form for all other dialects.
- Add TestBuildTypeProbeSQL (6 tests) to pin the dialect dispatch.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- docker-compose.yml: createdb service now uses mcr.microsoft.com/mssql/server:2022-latest image directly (has sqlcmd) instead of the custom Dockerfile that only has msodbcsql18 - Dockerfile/start.sh: fix stale --models-dir flag → --storage - ingestion.py: guard mssql.TIMESTAMP (rowversion/binary) from colliding with sa.TIMESTAMP in _SA_TYPE_MAP; isinstance check maps it to TEXT; add ROWVERSION entry for completeness - generator.py: add assert isinstance(iv, exp.Interval) guard in T-SQL DATEADD branch - test_sql_generator.py: add tsql to test_date_trunc_casts_unknown_typed_time_dim - test_ingestion.py: add test_tsql_mssql_timestamp_rowversion_maps_to_text Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… docs - generator.py: replace assert isinstance(iv, exp.Interval) with an explicit raise TypeError so the guard survives python -O optimization - client.py: add _ODBC_SQL_TYPE_MAP (standard ODBC SQL type codes) and route mssql/sqlserver/tsql integer type codes through it instead of falling through to the Postgres OID map; fixes get_column_types() and schema drift detection for SQL Server returning all-"string" types - test_sql_client.py: add parametrized tests covering SQL_INTEGER, SQL_BIGINT, SQL_VARCHAR, SQL_TYPE_TIMESTAMP, SQL_BIT, and the no-fallthrough invariant for all three T-SQL db_type aliases - docs/configuration/datasources.md: add warning that TrustServerCertificate=yes in auto-generated SQL Server connection strings is dev/Docker-only and must not be used in production Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…pin Dockerfile Debian - tests/test_sql_client.py: drop '= VALUE' from ODBC constant doc comments to avoid Sonar S125 "commented-out code" false positives - slayer/sql/generator.py: split shared MySQL/T-SQL percentile NotImplementedError into two separate dialect-specific messages for clearer diagnostics - examples/sqlserver/Dockerfile: pin base image to python:3.14-slim-bookworm so the msodbcsql18 apt install doesn't break when Debian 13 ships Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…redundant pip install - tests/test_sql_generator.py: add test_build_date_trunc_quarter to TestTsqlDialect to cover the previously untested QUARTER granularity path - docs/configuration/datasources.md: add SQL Server 2022+ version warning (DATETRUNC requires 2022+), clarify that auto-generated connection strings use pyodbc only (pymssql requires a manual connection_string) - examples/sqlserver/Dockerfile: remove redundant pip install of psycopg2-binary, pymysql, clickhouse-sqlalchemy — all already pulled in by poetry install -E all Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add parametrized tests for the four SQL Server extension type codes that were in _ODBC_SQL_TYPE_MAP but had no test coverage: - SQL_SS_TIMESTAMPOFFSET (-154) → time - SQL_SS_TIME2 (-155) → time - SQL_SS_XML (-152) → string - SQL_GUID / uniqueidentifier (-11) → string Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ion, not raise The module-level comment incorrectly said MySQL raises NotImplementedError for corr/covar functions. MySQL and T-SQL both use _build_covar_formula (variance-decomposition) instead. Update the comment to match. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The non-idempotent ingest_datasource was overwriting model metadata (descriptions, labels, allowed_aggregations) on every container restart. Switch to ingest_datasource_idempotent, which is additive-only and consistent with all other documented entry points. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
DATETRUNC/DATEADDfor date ops (noINTERVAL);STDEV/STDEVP/VAR/VARPfor 1-arg stat aggs; variance-decomposition formula forcorr/covar_samp/covar_pop(same formula now also used for MySQL — previously these raisedNotImplementedErroron MySQL);NotImplementedErrorformedian/percentile(PERCENTILE_CONTis window-only in T-SQL);iso_weekfor Monday-based week truncationDatasourceConfig:mssql/sqlserver/tsqltype aliases generatemssql+pyodbc://connection strings withODBC+Driver+18+for+SQL+Server&TrustServerCertificate=yes_SA_TYPE_MAPgains T-SQL-specific SQLAlchemy types (TINYINT,DATETIME2,SMALLDATETIME,DATETIMEOFFSET,NVARCHAR,NCHAR,NTEXT,MONEY,SMALLMONEY)pyproject.toml:pyodbc>=5.0optional dep +sqlserver = ["pyodbc"]extra; added toallexamples/sqlserver/: full Docker Compose example with a customDockerfilethat installsmsodbcsql18, SQL Server 2022 container, seed + verify scripts (coverscheck_stddev_var+check_corr_covar)examples/seed.py:CREATE_SQL_TSQLbranch usingNVARCHAR/DATETIME2(T-SQL doesn't haveTEXTorTIMESTAMPcolumn types)Test plan
poetry run pytest tests/test_sql_generator.py::TestTsqlDialect— 39 T-SQL-specific testspoetry run pytest tests/test_models.py::TestDatasourceConfig— connection string testspoetry run pytest tests/test_ingestion.py— T-SQL type mapping testspoetry run pytest -m "not integration"— 3516 passedcd examples/sqlserver && docker compose up -d && python verify.py🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Documentation
Tests
Chores