Add async support for Dataverse SDK by abelmilash-msft · Pull Request #171 · microsoft/PowerPlatform-DataverseClient-Python

abelmilash-msft · 2026-04-28T05:56:26Z

Introduces the complete aio/ async package mirroring the sync SDK:
- HTTP layer: _AsyncHttpClient wrapping aiohttp with identical retry, backoff, and timeout logic
- Auth: _AsyncAuthManager for async Azure Identity token acquisition
- OData client: _AsyncODataClient — full CRUD, SQL-over-API, table/column metadata, file upload, and relationship operations
- Batch: _AsyncBatchClient with _SyncResponseWrapper bridging the async HTTP response to the shared sync multipart parser in _BatchBase
- Operation namespaces: records, tables, query, files, batch, dataframe — all mirroring their sync counterparts
- Top-level client: AsyncDataverseClient with lazy init, async context manager, and session lifecycle management
Adds pytest-asyncio (asyncio_mode = auto) and aiohttp as an optional dependency (pip install PowerPlatform-Dataverse-Client[async]).

Test plan

1,758 unit tests pass (389 new async tests, ~95% coverage of the async layer)
black --check passes on all files

Copilot

Copilot wasn't able to review this pull request because it exceeds the maximum number of lines (20,000). Try reducing the number of changed lines and requesting a review from Copilot again.

- Reset to main (which is now PR #175) - Re-apply: _ODataBase, _BatchBase, _QueryBuilderBase, _BatchContext Protocol, _operation_context in base, Self type annotation - Re-export multipart helpers from _batch.py for test compatibility - Update test_sql_parse.py patch target to _odata_base.urlparse Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Reset to main via refactoring branch - Restore full async implementation: aio/ client, HTTP, batch, OData, relationships, upload, query builder, fetchxml, operations, tests, examples - Re-export multipart helpers from _batch.py for test compatibility - All 2166 tests passing Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Rename `r` to `response` for consistency with _async_http.py. Replace multi-pass comprehension + zip in _resolve_table_remove_columns with a plain for loop matching the sync _batch.py structure. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Replace broad try/except around r.json() with hasattr guard, matching the sync FetchXmlQuery pattern. Remove the now-obsolete test that verified the exception-swallowing behavior. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Moves fetchxml() after sql_columns() to match the order in the sync QueryOperations: builder, sql, fetchxml, sql_columns, odata_*. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Order now matches sync exactly: builder, sql, fetchxml, sql_columns, odata_expands. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Add missing "Use for SQL-JOIN scenarios..." sentence, expand :return: description, add link-entity JOIN example, align Eager/Lazy comment wording, and add the two inline implementation comments from sync. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…ession on auth failure - functional_testing.py: replace batch.records.get() with batch.records.retrieve(); close client session on auth failure to avoid unclosed session warning - sql_examples.py: remove odata_select/odata_expand/odata_bind calls (deprecated sync-only helpers not available on async client); keep only odata_expands() Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…ontent _AsyncResponse buffers the body in ._body; it has no .content attribute (unlike requests.Response used by the sync client). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…replace downloads Two download paths after replace uploads still used resp_r.content and resp_rc.content instead of ._body, causing AttributeError. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Validates 7 properties of the async client against a live environment: 1. Non-blocking reads — canary confirms event loop stays free during GETs 2. Read throughput — concurrent reads via gather() beat sequential 3. Write concurrency — concurrent POSTs beat sequential (POST path) 4. Pagination non-blocking — async generators yield between page fetches 5. Mixed fan-out — different op types run simultaneously without serialization 6. Error resilience — one failure in gather() does not kill other calls 7. Real-world fan-out — metadata for multiple tables fetched in parallel Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Each test now has a brief explanation of what it runs, what property it validates, and what a failure would indicate. Also clarifies that speedup measures async-sequential vs async-concurrent, not async vs sync. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

abelmilash-msft force-pushed the users/abelmilash/async-phase2 branch from e491241 to b848324 Compare May 11, 2026 23:53

abelmilash-msft marked this pull request as ready for review May 12, 2026 06:29

abelmilash-msft requested a review from a team as a code owner May 12, 2026 06:29

Copilot AI review requested due to automatic review settings May 12, 2026 06:29

Copilot AI reviewed May 12, 2026

View reviewed changes

abelmilash-msft force-pushed the users/abelmilash/async-phase2 branch from b32987a to ad9297e Compare May 14, 2026 17:28

github-actions Bot reviewed May 15, 2026

View reviewed changes