fit_all_damping_curves(save_txt=True) checks func_serialize and txt_filename only after all the fitting is done #97

Found while reviewing #74. Pre-existing on main (b764057); #74 keeps the same behavior.

Problem

fit_all_damping_curves() fits every layer first, and only then looks at the save_txt options. A missing func_serialize raises ValueError after the fitting. A missing txt_filename (the default is None) is not checked at all: np.savetxt(None, ...) fails after the fitting. With the default pop_size=800 and n_gen=100, the fitted parameters are lost after a long run.

import numpy as np
import PySeismoSoil.helper_hh_model as hh
import PySeismoSoil.helper_site_response as sr

curve = np.genfromtxt('tests/files/curve_FKSH14.txt')[:, 2:4]
n_fits = 0

def counting_fit(*args, **kwargs):
    global n_fits
    n_fits += 1
    return hh.fit_HH_x_single_layer(*args, **kwargs)

for kwargs in (
    {'func_serialize': None},
    {'func_serialize': hh.serialize_params_to_array},  # txt_filename=None
):
    n_fits = 0
    try:
        sr.fit_all_damping_curves(
            [curve] * 3, counting_fit, hh.tau_HH,
            pop_size=5, n_gen=5, save_txt=True, **kwargs,
        )
    except ValueError as e:
        print(n_fits, e)
# 3 Please provide a function to serialize the parameters into a list.  (expected 0 fits)
# 3 fname must be a string or file handle  (expected 0 fits, and a clear message)

The existing test test_fit_all_damping_curves__exception_when_no_func_serialize on main (in #74: test_fit_all_damping_curves__exception[no_func_serialize]) only checks that the ValueError is raised, not when.

Also:

• save_fig=True saves nothing unless show_fig=True too, because the figure is only built, and only saved, when show_fig is True (if show_fig and save_fig:). The docstring of save_fig does not say this. Multiple_Damping_Curves.get_all_HH_x_params() and get_all_H4_x_params() pass save_fig through, so they also save nothing when show_fig=False.
• show_fig=True, save_fig=True with the default fig_filename=None fails after the fitting, inside fig.savefig(None).

import os
import tempfile
import numpy as np
import PySeismoSoil.helper_hh_model as hh
import PySeismoSoil.helper_site_response as sr
from PySeismoSoil.class_curves import Multiple_Damping_Curves

curve = np.genfromtxt('tests/files/curve_FKSH14.txt')[:, 2:4]
fig_file = os.path.join(tempfile.mkdtemp(), 'fit.png')
sr.fit_all_damping_curves(
    [curve], hh.fit_HH_x_single_layer, hh.tau_HH,
    pop_size=1, n_gen=1, save_fig=True, fig_filename=fig_file,
)
print(os.path.exists(fig_file))  # False  (expected True)

mdc = Multiple_Damping_Curves('tests/files/curve_FKSH14.txt')
mdc.get_all_HH_x_params(pop_size=1, n_gen=1, save_fig=True, fig_filename=fig_file)
print(os.path.exists(fig_file))  # False  (expected True)
mdc.get_all_H4_x_params(pop_size=1, n_gen=1, save_fig=True, fig_filename=fig_file)
print(os.path.exists(fig_file))  # False  (expected True)

sr.fit_all_damping_curves(
    [curve], hh.fit_HH_x_single_layer, hh.tau_HH,
    pop_size=1, n_gen=1, show_fig=True, save_fig=True,
)
# AttributeError: 'NoneType' object has no attribute 'write'

Fix

1. At the start of fit_all_damping_curves(), before any fitting, raise ValueError when:
   • save_txt is True and func_serialize is None (move the existing check here);
   • save_txt is True and txt_filename is None;
   • save_fig is True and fig_filename is None.
2. Build the figure when show_fig or save_fig, and save it when save_fig (instead of show_fig and save_fig).
3. Tests: add {'func_serialize': hh.serialize_params_to_array} (no txt_filename) as a ValueError case to test_fit_all_damping_curves__exception (Combine near-duplicate test functions into parametrized tests #74), and pass a fitting function that fails the test if it is called, so the cases check that no fitting runs. Add a test that save_fig=True, fig_filename=tmp_path / 'fit.png' writes the file with show_fig=False.

With only steps 1 and 2 on main, both repros above give the expected results (0 fits before each error, and the figure files are written), and all 240 tests pass.