fix: close file handles via context managers, fix Config string-concat and mutable default arg bugs

[suyash469]

Summary

Fixes three real bugs found in the API base classes during code review.

What changed

1. FileClass.py — Resource leak: unclosed file handles

All four methods (readFile, writeFile, writeFileUJson, readParamFile) used
bare open() + close(). If an exception fired between them, the file handle was
never closed — causing "Too many open files" OS errors under load and potential
data corruption on writes.

Fix: Replaced all four open/close pairs with with context managers, which
guarantee closure even on exception.

2. Config.py — Silent string concatenation in PARAMETERS_C

A missing comma between 'e' and 't' in the EmissionActivityRatio entry:

# Before (bug): 'e''t' is silently parsed by Python as 'et' → 4 dims
'EmissionActivityRatio': ['r', 'e''t', 'y', 'm'],

# After (fix): correct 5 dimensions
'EmissionActivityRatio': ['r', 'e', 't', 'y', 'm'],

[NamanmeetSingh]

Hi Suyash, great catch on the mutable default argument (kwargs={}) in the CustomThread class! That’s a subtle Python gotcha that can definitely cause state-leakage bugs.

Also, since you are refactoring FileClass.py, I’ll drop the an optimization note here:
In readFile and readParamFile, keeping json.loads(f.read()) pulls the entire text payload into memory before parsing. For standard configs this is fine, but as we prep for the OG-CLEWS integration, we will be loading massive, multi-country scenario files. Switching these to json.load(f) will allow the parser to stream directly from the file, preventing massive RAM spikes during the model runs.

Really clean fixes overall, especially the thread state correction.

[MahadevBalla]

This is a good cleanup - switching to context managers definitely fixes the file handle leak issue but one thing I noticed is that we’re still opening files with "w", which truncates the file immediately. So if something crashes mid-write (disk full, process killed, etc.), the file could still end up partially written or empty.

Maybe it’s worth considering a temp-file + atomic rename approach here? That way the original file isn’t touched unless the new write fully succeeds. Since these JSON files hold core model data, that might make the writes a bit more resilient.

[suyash469]

Hi Suyash, great catch on the mutable default argument (kwargs={}) in the CustomThread class! That’s a subtle Python gotcha that can definitely cause state-leakage bugs.

Also, since you are refactoring FileClass.py, I’ll drop the an optimization note here: In readFile and readParamFile, keeping json.loads(f.read()) pulls the entire text payload into memory before parsing. For standard configs this is fine, but as we prep for the OG-CLEWS integration, we will be loading massive, multi-country scenario files. Switching these to json.load(f) will allow the parser to stream directly from the file, preventing massive RAM spikes during the model runs.

Really clean fixes overall, especially the thread state correction.

Great point on json.load(f) vs json.loads(f.read()) — the streaming approach
is clearly better for large OG-CLEWS scenario files. Will update readFile and
readParamFile to use json.load(f) directly.