Update test_network.py

[bosd]

No description provided.

[bosd]

Triage update

The substance here is not obsolete — the parser changes are legitimate:

• compute_plausible_gaps: h_textlines[i].x0 - h_textlines[i-1].x1 (gap between text edges) replaces x0 - x0 (gap between text starts). The new formula is the right one for column-spacing detection — text-width-dependent biases disappear.
• max(..., default=0) on the merged-zones loop in search_header_from_body_bbox is a real safety fix for the empty-iterator case.
• New good_energy.pdf fixture is a genuine reproducer for Unable to specify column separators for only one table when there are multiple tables #585-style column-separator confusion.

Why CI is fully red:

• pre-commit 3.10 fail = branch predates a couple of pre-commit-config.yaml / black / flake8 changes that landed since June.
• Every test row failing in lockstep = the network gap-formula change shifts what counts as a column boundary on the existing test fixtures (foo.pdf, budget.pdf, …). Either those fixtures need their pinned column/row coords updated, or the formula change needs to be opt-in.

Action items if you want to resurrect this:

1. Rebase onto current master (mechanical).
2. Run tests/test_network.py locally and decide for each diff between old and new output: is the new output correct (then update the fixture)? Or did we lose information (then narrow the formula change)?
3. Add a test specifically exercising good_energy.pdf (the regression that prompted the PR).

I won't auto-rebase or close — this is your work and the parser side has nuance. Happy to do the rebase + fixture update pass if you confirm the direction; just say so. Leaving open for now.

[bosd]

Closing as superseded by a clean 3-PR split:

• fix(network): defensive guards in TextNetworks / Network (#619 split 1/3) #744 (merged) — the two safe defensive guards in TextNetworks (return None on missing connected textline, max(..., default=0) on empty zones).
• test(network): #585 fixture + two xfailed regression tests (#619 split 2/3) #745 (merged) — the good_energy.pdf fixture + two xfailed regression tests pinned to Unable to specify column separators for only one table when there are multiple tables #585's specific symptom.
• 3/3 (not yet opened) — the gap-edge formula change (x0 - x1 instead of x0 - x0). Held back because it shifts the column-spread math in search_header_from_body_bbox and needs the xfailed tests of test(network): #585 fixture + two xfailed regression tests (#619 split 2/3) #745 to flip to xpass before it can be merged with confidence.

The defensive bits no longer need this branch as their staging ground; closing for hygiene. Will reopen or supersede with a new PR for 3/3 once the local validation against the #585 reproducer is done.