Defer pandas/omegaconf imports

[ptomecek]

Summary

Defer eager imports of pandas and omegaconf so they no longer load on a bare import ccflow. Pure import-site relocation; runtime behaviour unchanged.

Changes

• ccflow/exttypes/frequency.py: drop module-level import pandas as pd and from pandas.tseries.frequencies import to_offset. Re-import locally in Frequency.offset, Frequency.timedelta, and Frequency._validate. Replace the module-level _PD_GE_22 constant (which read pd.__version__ at import) with a @cached _pd_ge_22() function called from _normalize_frequency_alias.
• ccflow/utils/chunker.py: move import pandas as pd into dates_to_chunks.
• ccflow/validators.py: move import pandas as pd into normalize_date and normalize_datetime.
• ccflow/base.py: drop module-level import omegaconf and from omegaconf import DictConfig, OmegaConf. Use if TYPE_CHECKING: from omegaconf import DictConfig so static type-checkers still resolve the annotations, and quote the runtime DictConfig annotations on ModelRegistry.load_config, ModelRegistry.create_config_from_path, and the _ModelRegistryLoader methods. Add local imports of DictConfig/OmegaConf/omegaconf inside _is_config_subregistry, _make_subregistries, and _ModelRegistryLoader.load_config where they are actually used at runtime.

Measurements

Profiled in a clean Python 3.11 venv with only ccflow's base deps installed.

	Before	After
import ccflow total	~525 ms	~320 ms (~38% faster)
pandas in sys.modules after import	yes	no
omegaconf in sys.modules after import	yes	no
antlr4 in sys.modules after import	yes (via omegaconf)	no

Compatibility

• All public re-exports from ccflow (e.g. from ccflow import Frequency) still work.
• No type-checking break: the quoted "DictConfig" annotations plus the TYPE_CHECKING import keep static type information intact.
• Smoke-tested: Frequency validation (including legacy alias path that exercises _pd_ge_22()), normalize_date, normalize_datetime, and dates_to_chunks all still work.

[github-actions]

Test Results

729 tests  ±0   727 ✅ ±0   1m 43s ⏱️ -7s
  1 suites ±0     2 💤 ±0 
  1 files   ±0     0 ❌ ±0 

Results for commit 22e4ca4. ± Comparison against base commit 59bc0e3.

♻️ This comment has been updated with latest results.

[codecov]

Codecov Report

❌ Patch coverage is 90.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.33%. Comparing base (60ee2e7) to head (22e4ca4).
⚠️ Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
ccflow/base.py	83.33%	1 Missing ⚠️
ccflow/exttypes/frequency.py	90.90%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #207      +/-   ##
==========================================
+ Coverage   95.22%   95.33%   +0.10%     
==========================================
  Files         140      142       +2     
  Lines       10354    10714     +360     
  Branches      599      613      +14     
==========================================
+ Hits         9860    10214     +354     
- Misses        369      374       +5     
- Partials      125      126       +1     

☔ View full report in Codecov by Sentry.
📢 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.