Convert XLS cell values when importing a Databook

[vineethsaivs]

The bug

XLSFormat.import_set reads cells through a cell_value helper that turns Excel date serials into datetime objects and error codes into strings like #N/A. XLSFormat.import_book re-implemented the row loop with raw sheet.row_values(i) and never got that helper, so the same file returns different values depending on whether you load it as a Dataset or a Databook.

Using the fixtures already in the test suite:

dates.xls    'birth_date'
   Dataset  -> datetime.datetime(2015, 4, 12, 0, 0)
   Databook -> 42106.0

errors.xls   'not available (formula)'
   Dataset  -> '#N/A'
   Databook -> 42

That 42 is the symptom from #202 ("Tablib import Excel NA() values as 42"), which was fixed on the import_set path only; it still reproduces today through Databook.

XLS is the only spreadsheet format where the two paths differ. Both XLSX and ODS delegate import_set and import_book to a shared import_sheet, so they cannot drift. The existing test_xls_date_import and test_xls_import_with_errors assert the converted values, but only via Dataset, and the one XLS Databook test only checks a sheet title, which is why this went unnoticed.

The fix

Extract an import_sheet classmethod and call it from both entry points, mirroring the structure _xlsx.py and _ods.py already use. The cell conversion becomes a read_cell classmethod, matching the read_cell in _ods.py, with the workbook datemode passed in explicitly rather than captured from an enclosing scope (sheet.book is not reliable under xlrd's on_demand).

Header handling is deliberately unchanged. import_set reads headers with raw row_values today and the shared helper keeps doing that, so Databook now matches Dataset exactly rather than changing both.

Tests

Two regression tests load the existing dates.xls and errors.xls fixtures as a Databook and assert the same values the Dataset tests already assert. Both fail before this change (42106.0 != datetime.datetime(2015, 4, 12, 0, 0)) and pass after.

$ pytest tests/ -q
182 passed

Full suite passes before and after; the count goes from 180 to 182 with the two new tests. Added myself to AUTHORS per the contributing guide.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.47%. Comparing base (234f434) to head (4189fab).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #665      +/-   ##
==========================================
+ Coverage   93.42%   93.47%   +0.05%     
==========================================
  Files          29       29              
  Lines        3298     3308      +10     
==========================================
+ Hits         3081     3092      +11     
+ Misses        217      216       -1     

☔ View full report in Codecov by Harness.
📢 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.

[vineethsaivs]

The only red check here is label: "Label error. Requires at least 1 of: changelog: Added, changelog: Changed, ...". I do not have rights to label the PR myself, so could a maintainer add changelog: Fixed? Everything else (tests, codecov) is green.

Recap: Databook import goes through the sheet rows directly, so XLS cell values are never passed through the same conversion the single-Dataset path applies. Dates, times and booleans therefore come back as raw floats/ints in a Databook but as proper Python objects in a Dataset from the same file. The fix routes both paths through the same conversion, with a regression test covering the mismatch.

[claudep]

Looks good, thanks for this fix!