[saurabh-saraswat]

Description

Added support for parameter exclude_attribute_keys in View. During consume_measurement removed the attributes which were in exclude_attribute_keys

Fixes #5331

Type of change

Please delete options that are not relevant.

• [✔] Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• [ ✔ ] Use exclude parameter while creating view like 👍

  view = View(
            instrument_name="api_counter",
            exclude_attribute_keys={"email", "phone"},
        )

Does This PR Require a Contrib Repo Change?

No

• [ ✔] Yes. - Link to PR: Add support for excluding attributes in metrics View #5332
• No.

Checklist:

• [ ✔] Followed the style guidelines of this project
• [✔ ] Changelogs have been updated
• [✔ ] Unit tests have been added
• Documentation has been updated

[xrmx]

Looks like you haven't run the tests locally

[saurabh-saraswat]

@xrmx . I have updated the pull request for the comments provided. Local tests was passing.

[Copilot]

Pull request overview

Adds View-level support for excluding specific attribute keys from metric stream identity, aligning the Python SDK View API with the Metrics SDK spec’s exclude-list behavior (issue #5331).

Changes:

• Introduces exclude_attribute_keys on View and applies it during measurement consumption to remove matching keys before stream identification.
• Updates declarative meter provider configuration to map attribute_keys.excluded into View(exclude_attribute_keys=...).
• Adds/updates unit tests and changelog entry for the new behavior.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
opentelemetry-sdk/src/opentelemetry/sdk/metrics/_internal/view.py	Adds exclude_attribute_keys to the View API and stores it for later use.
opentelemetry-sdk/src/opentelemetry/sdk/metrics/_internal/_view_instrument_match.py	Applies the exclude-list while building the attribute set used for aggregation keying.
opentelemetry-sdk/src/opentelemetry/sdk/_configuration/_meter_provider.py	Wires declarative config’s excluded attribute keys into the SDK View.
opentelemetry-sdk/tests/metrics/test_view_instrument_match.py	Adds tests validating exclude-list behavior and include-then-exclude precedence.
opentelemetry-sdk/tests/_configuration/test_meter_provider.py	Updates config test to assert excluded attribute keys are applied to View.
.changelog/5332.added	Documents the added View(exclude_attribute_keys=...) support.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[saurabh-saraswat]

Updated for github review comments

[saurabh-saraswat]

I have incorporated comments, let me know if something is pending from my side

[saurabh-saraswat]

@herin049 thanks for comments, I have incorporated the changes. Kindly check and help in closure of pull request.

[saurabh-saraswat]

@herin049, @xrmx I have resolved all the comments. Kindly go through it and see if this pull request can be merged now.

[herin049]

There are several issues here identified by Codex:

[P1] Preserve attributes when filtering is unconfigured.
In the final filtering branch, attributes = {} discards the copied attributes when neither list is configured. This also affects default streams with no registered views.

Reproduction: recording two counter measurements with region="us" and region="eu" produces one attribute-free point with value 2, instead of two distinct points. An explicitly empty exclusion list has the same problem.

The review should request retaining the copied attributes when no filter applies.

[P2] Pattern-string matching does not establish that patterns are disjoint.
The constructor’s overlap check tests whether either pattern string matches the other. That misses patterns whose matching sets intersect.

Reproduction: inclusion http.* and exclusion *.status pass construction, although both match http.status. Likewise, a? and ?b both match ab. Because measurement filtering uses elif, accepted configurations silently ignore the exclusion whenever inclusion is configured.

The review should distinguish this from valid exact-overlap rejection: failing fast is allowed, but this check cannot guarantee disjointness. Recommend accepting intersections and applying exclusions after inclusions, avoiding a glob-intersection algorithm.

[P2] Direct fnmatchcase calls violate literal and declarative matching behavior.
The attribute matching expressions interpret bracket expressions as character classes.

Reproduction with both header[0] and header0 present:

• Including header[0] retains header0 and drops the identical literal key.
• Excluding header[0] removes header0 and retains the identical literal key.

This breaks existing literal-key inclusion behavior and the declarative schema’s matching contract. The constructor also falsely rejects inclusion header0 plus exclusion header[0].

Request exact matching plus wildcard handling limited to * and ?, treating brackets and backslashes literally.

[P2] Declarative configurations cannot express the specified exclusion precedence.
The configuration adapter forwards both lists to the rejecting constructor.

Reproduction: included=["region"], excluded=["region"] raises ValueError, rather than constructing a stream that excludes region. Inclusion http.* plus exclusion http.status also raises.

SDK-level permission to reject conflicts does not implement the declarative schema’s explicit exclusion-priority behavior. Recommend applying exclusion precedence consistently in both interfaces