Identify a PDF table with nested subrows

Where the extra work starts

Decide how the hierarchy should be represented so extraction produces something usable for your next calculation or review.

For supplier reports and technical tables, start with ParseMill’s synthetic material-register example. Compare the source references and output with the register you need, then discuss a document pilot with an agreed scope and fee. The self-service trial covers invoice PDFs.

A blank parent field in a subrow is not necessarily missing data. It may mean the value is inherited visually from a group, but that relationship must be supported by the source rather than assumed.

A practical way to get it right

Read a complete group from its heading to the next group boundary. Identify parent labels, child rows, subtotals and notes. For each child, record which context applies and how the source expresses it: indentation, spanning cells, repeated labels or explicit references.

Build a manual hierarchy map with parent reference, child reference and source location. If a flat register is required, define how parent context will be represented in separate fields and have the reviewer confirm the rule. Preserve the original relationship so another user can reconstruct the group.

Choose a method that can maintain the required hierarchy. ParseMill's invoice workflow does not promise general nested-subrow interpretation for report tables. A bounded trial must therefore be evaluated against the relationship map, with unsupported or ambiguous groups retained as exceptions.

W3C: PDF table markup distinguishes table rows, cells and spanning relationships. Inspect those relationships before treating visually nested content as ordinary independent rows.

What this looks like on a real task

Synthetic example. The people, references and figures below are invented to demonstrate the method.

The source has two test records under S8, not three unrelated data rows including the parent label. Copy the visual-row and hierarchy columns, then add explicit parent references to any reviewed flat output. The indentation shown is synthetic; your own source may communicate grouping differently.

Catch the cases that create rework

Blind fill-down can attach a child to the wrong parent after a page break or subtotal. Verify every group boundary and inspect continuation pages. Do not infer a parent from the nearest nonblank value if a note or section change intervenes.

A parent row can also contain its own data, so it should not automatically be discarded as a heading. Record its role from the source and the intended output definition.

W3C: PDF reading order explains that reading order depends on PDF structure. A copied text sequence alone may not retain the parent-child context visible on the page.

Decide whether the change is worth keeping

Ask a reviewer to reconstruct the original groupings from the proposed flat register. Every child should retain the correct parent and any relevant condition, while summary rows remain distinct. If that cannot be done reliably, preserve the hierarchy in a more suitable representation or use reviewed manual transcription rather than accepting a lossy flattening.