Base Import Pdf by Template - Engine

Two small extension points on base_import_pdf_by_template's models, factored out so a third-party module can plug in a different pattern language (JSONPath, ...) or build tooling around pattern matching (a live preview, ...) without overriding - and duplicating - whole methods from the base module.

Why

The base module hardcodes regex in three places: table-column extraction (_get_table_info_data), header extraction (_get_field_value), and value post-processing (_process_value). The first two are easy to override per line. The third is not, because a pattern engine whose selection step already consumes the pattern (a JSONPath expression finds its match while extracting the value, unlike a regex which is re-applied inside _process_value on top of an already-extracted string) still needs _process_value's ~25-line tail - date/float conversion, mapped_ids, search_field_id, default_value - without re-running the pattern a second time.

This module adds two seams instead of solving that per consumer module:

• base.import.pdf.template.line._get_column_values(text) - the values this line contributes to a "lines" table, as a column. base.import.pdf.template._get_table_info_data() is refactored to build each column by calling this per child line instead of inlining a regex loop, so overriding one line's extraction no longer means overriding table assembly too. As a side effect this also fixes two latent bugs in the base algorithm: an IndexError when no child line has a pattern yet, and a silent column-shift when one column has fewer matches than another (a short column is now padded with "" rather than dropped, which keeps every later column of that row aligned with the right field).

• base.import.pdf.template.line._without_pattern() - an in-memory twin of a line with pattern cleared, safe to call super()._process_value(value) on. Correctly falls through to the real record's values whether self is a saved record, or is itself already an in-memory/NewId record (an onchange, a live preview compute, or odoo.tests.Form) - the latter needs an explicit value copy rather than new(origin=self), because a NewId origin of a NewId is silently dropped by odoo.models.origin_ids(), which would otherwise make every stored field on the twin fall back to its bare field default instead of the real value.

Neither seam introduces a new field, a new model, or a UI change on its own - see import_via_xberg (a JSONPath pattern engine) and import_preview (a live match preview) for modules that build on top of them.

Tests

python odoo/odoo-bin -c odoo.conf -d <testdb> -i base_import_pdf_by_template_engine \
  --test-enable --test-tags /base_import_pdf_by_template_engine --stop-after-init