[MohammadHijjawi97]

What this PR does / why we need it:

The ibis point-in-time join (used by the DuckDB, MSSQL and Oracle offline stores) identifies each entity row by concatenating the string values of its join keys and event timestamp into entity_row_id, then deduplicates and joins feature rows back on that id. Two problems with plain concatenation:

1. Different entity rows can get the same id, e.g. driver_id="1", customer_id="11" and driver_id="11", customer_id="1" at the same timestamp both become "_111...", so one row silently receives the other row's feature values (whether it happens depends on the column order, which comes from a set).
2. A null key makes the whole id null, so the row loses the features of every feature view, not just the one keyed on the null column.

Repro with point_in_time_join (drivers FV keyed on driver_id, customers FV keyed on customer_id):

Before:
    driver_id customer_id  driver_rating  customer_score
            1          11            1.0           110.0
           11           1            1.0           110.0   <- driver 1's values
            7        None            NaN             NaN   <- driver_rating lost
After:
            1          11            1.0           110.0
           11           1           11.0            10.0
            7        None            7.0             NaN

This length-prefixes each value and gives nulls their own marker, so the id is only shared by rows with equal values. Both places that build the id now use the same helper.

Which issue(s) this PR fixes:

No existing issue.

Checks

• I've made sure the tests are passing.
• My commits are signed off (git commit -s)
• My PR title follows conventional commits format

Testing Strategy

• Unit tests
• Integration tests
• Manual tests
• Testing is not required for this change

Misc

Added test_point_in_time_join_keeps_entity_rows_apart to sdk/python/tests/unit/infra/offline_stores/test_ibis.py; it fails on master (deterministically, through the null-key case) and passes with this change. tests/unit/infra/offline_stores/test_filter_by_created_timestamp.py (the other caller of point_in_time_join) gives the same results as on master. ruff and mypy on ibis.py are clean. The id format is internal and dropped before results are returned.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 88.88889% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 47.84%. Comparing base (ce37029) to head (881cc39).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
sdk/python/feast/infra/offline_stores/ibis.py	88.88%	1 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #6884   +/-   ##
=======================================
  Coverage   47.83%   47.84%           
=======================================
  Files         422      422           
  Lines       52727    52727           
  Branches     7663     7662    -1     
=======================================
+ Hits        25223    25225    +2     
+ Misses      25683    25680    -3     
- Partials     1821     1822    +1     

Flag	Coverage Δ
go-feature-server	30.58% <ø> (ø)
python-unit	49.18% <88.88%> (+<0.01%)	⬆️

Files with missing lines	Coverage Δ
sdk/python/feast/infra/offline_stores/ibis.py	41.62% <88.88%> (+1.43%)	⬆️

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ce37029...881cc39. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.