fix(mfusg): fix precision loss in CLN/GNC packages writing coordinates

[reneangermeyer]

MFUSG CLN/GNC packages were losing significant precision when writing and affecting model results.

Root Cause

The fmt_string() method in mfusgcln.py and mfusggnc.py used %10.2e format (only 3 significant figures) instead of the standard Util2d format %15.6G (9 significant figures).

Impact on Real Data

Complete line from CLN package showing precision loss:

Before (bug - %10.2e):
1 1 0 5.75e+01 1.15e+03 0.00e+00 0 0 4.01e+05 6.90e+06 1.21e+03 4.01e+05 6.90e+06 1.15e+03

After (fixed - %16.9G):
1 1 0 57.4650002 1150.19897 0 0 0 400569 6903567 1207.66394 400569 6903567 1150.19897

The precision loss caused measurable differences in model results.

Solution

Changed fmt_string() to return "%16.9G" format, which:

• Provides 9 significant figures (exceeds float32's 7-digit precision)
• Uses adaptive format (decimal when readable, scientific when needed)
• Maintains proper spacing with 16-character width
• Consistent with other MFUSG packages (WEL/DRN/RIV use %G)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 72.6%. Comparing base (556c088) to head (2c85e71).
⚠️ Report is 90 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2658      +/-   ##
===========================================
+ Coverage     55.5%    72.6%   +17.1%     
===========================================
  Files          644      667      +23     
  Lines       124135   129360    +5225     
===========================================
+ Hits         68947    93990   +25043     
+ Misses       55188    35370   -19818     

Files with missing lines	Coverage Δ
flopy/mfusg/mfusg.py	78.3% <100.0%> (+0.8%)	⬆️
flopy/mfusg/mfusgcln.py	69.2% <100.0%> (+2.2%)	⬆️

... and 557 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[wpbonelli]

@reneangermeyer the USG executable seems not to like something about the new format?

[Copilot]

Pull request overview

This PR fixes a critical precision loss bug in MFUSG's CLN and GNC packages by updating the float format specifier from %10.2e (only 3 significant figures) to %16.9G (9 significant figures). This precision loss was causing measurable differences in model results.

• Changed the format string for float values in the shared fmt_string() function
• The fix applies to both CLN and GNC packages which import this function
• Increases precision from 3 to 9 significant figures, exceeding float32's 7-digit precision

---

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

[reneangermeyer]

@wpbonelli The issue was that the test model doesn't use FREE format, so %16.9G violates the 10-character field width limit required by MODFLOW-USG.

However, the precision loss with %10.2e is still a real problem (values like 57.465 become 5.75e+01) which can affect model results. The fix now checks if FREE format is enabled in the BAS file: if so, it uses %16.9G for full precision, otherwise, it falls back to the original %10.2e to maintain compatibility with fixed-width input.

[wpbonelli]

Ah, got it. Seems reasonable.