fix(archive): anchor component name when matching history files

[chengzhuzhang]

Problem

ArchiveBase.get_all_hist_files() builds its match regex by concatenating the component name directly onto the extension pattern:

string = model + r"\d?_?(\d{4})?(_d\d{2})?\." + ext

and then uses re.search, so the component name matches anywhere in the filename — including inside another component's name. The concrete case is E3SM's atmosphere components: eam is a substring of scream, so the eam spec matches EAMxx history files:

>>> archiver.get_all_hist_files("casename", "eam", rundir)
['casename.scream.h.AVERAGE.nmonths_x1.0001-01.nc']   # not an eam file

casename.scream.h... contains eam (in scr-eam) followed by ., and eam's extension h\d*.*\.nc$ then matches the rest.

In a real case only one atmosphere spec is loaded, so this does not corrupt production archiving. Where it does bite is anywhere several specs are in play at once — test_st_archive, which loads every comp_archive_spec in the model — so a component's spec can silently "cover" for another component's broken or missing pattern, and the test passes when it should not. That is exactly what happened in E3SM: a wrong hist_file_extension for scream went unnoticed because eam's pattern matched the files instead (E3SM-Project/E3SM#6190, E3SM-Project/E3SM#8589).

Fix

Anchor the component name at a filename boundary — start of name, or preceded by ., -, or _:

string = r"(?:^|[\.\-_])" + model + r"\d?_?(\d{4})?(_d\d{2})?\." + ext

The leading ^ alternative keeps the existing f.startswith(model) case working, where a file is named after the component rather than the case. The added group is non-capturing, so the existing (\d{4})? / (_d\d{2})? group numbering is unchanged (nothing reads those groups here, but it keeps the diff behavior-neutral).

Every separator that appears before a component name in the naming conventions I could find is covered: casename.eam.h0…, casename.eam_0002.e…, casename.clm2_0002.h0…, ga0xnw.mom6.frc…, casename.mpaso.hist.am…, and the long test-name form …C.fake_testing_only_20160816_164150-20160816_164240.cpl.h.nc.

Tests

Two new tests in CIME/tests/test_unit_xml_archive_base.py, written as pytest functions per AGENTS.md:

• test_component_name_is_not_matched_as_substring — with both an eam and a scream spec loaded, each component returns only its own history file. Fails on master, passes with this change.
• test_component_name_at_start_of_filename — covers the ^ alternative, i.e. a file named after the component instead of the case.

Verified locally:

• pytest CIME/tests/test_unit_xml_archive_base.py CIME/tests/test_unit_case_st_archive.py — 14 passed.
• pytest CIME/tests/ -k unit — 486 passed, 15 skipped, 3 failed; the same 3 (test_unit_doctest::test_lib_docs and two test_unit_system_tests::test_generate_baseline*) fail identically on unmodified master in this environment, so they are pre-existing and unrelated.
• black --check clean on both files.

Note

_get_extension() in the same module builds its regex the same way and has the same substring exposure. I left it alone to keep this change small, since it is only reached with a single component's model name at a time; happy to extend it here if you would prefer both fixed together.

Analysis and patch produced by Claude (Claude Code).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 28.65%. Comparing base (f8e59ff) to head (31474a7).
⚠️ Report is 104 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5021      +/-   ##
==========================================
+ Coverage   28.48%   28.65%   +0.16%     
==========================================
  Files         262      266       +4     
  Lines       38490    39024     +534     
  Branches     8146     8278     +132     
==========================================
+ Hits        10964    11181     +217     
- Misses      26268    26568     +300     
- Partials     1258     1275      +17     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.