`report_sections`

Committee report heading and text blocks

One row per heading or unheaded text block in one part of a committee report, keyed (package_id, part_id, seq). heading preserves the source title; agency identity fields remain unresolved. pattern records which header pattern fired, which is this table's provenance column. All columns are stored as VARCHAR.

Coverage. Sampled. Heading/text blocks of the reports in committee_reports. heading retains the source title while agency_label and agency_key remain unresolved NULLs. HTML bodies carry no page boundaries, so none are inferred. (measured 2026-09-28)

Data quality. part_id names the committee_reports row a block was parsed from, so (package_id, part_id) is its parent, and seq counts from zero within that part. By decision 29 (docs/research/fork-delivery-decisions-2026-09-22.md) part_id is the package id for a report published in one part and the part's granule id for a report published as parts, CRPT-119hrpt494's unsuffixed Part 1 aside; a package whose one part is spelled -pt1 (CRPT-119hrpt811; spicy-docs docs/tables.md names the others it measured) carries that spelling, so the parent's row count tells a lone part from Part 1 of several. A block published before part_id existed carries its package id there until its report is read again, which replaces every block of the package.

Parquet file: report_sections.parquet
MCP query_sql support: Configured; requires an available artifact.
Publication status: Not established by this schema page or its measurement date.
Row count: Not stated here; the MCP describe_table reply gives the live count under publication.

Column	Type	Description
`package_id`	`VARCHAR`	The report package this block was parsed from.
`seq`	`VARCHAR`	Zero-based position of this block in its part, in reading order.
`agency_label`	`VARCHAR`	NULL because heading recognition does not establish agency identity; the source heading is retained separately in heading.
`agency_key`	`VARCHAR`	NULL because this reader does not resolve headings to agency identities.
`body`	`VARCHAR`	The block's text, trimmed the way the original trimmed it.
`pattern`	`VARCHAR`	Which header pattern fired to start this block; this table's provenance column. Two values are not pattern names: `preamble` for the text before the first header, and `full_report` for a report where no header matched at all.
`char_start`	`VARCHAR`	Start offset of this block's span in the flattened report text.
`char_end`	`VARCHAR`	End offset, exclusive; spans partition the whole input with no gap and no overlap.
`page_start`	`VARCHAR`	First page the span touches, where the input carried page boundaries.
`page_end`	`VARCHAR`	Last page the span touches, where the input carried page boundaries.
`body_chars`	`VARCHAR`	Character length of body, which can be shorter than the span it sits in.
`last_modified`	`VARCHAR`	The parent report's last_modified, carried so this table versions with the package it came from.
`heading`	`VARCHAR`	The source heading as spelled and trimmed, including actual agency names and generic titles; NULL for preamble or full_report blocks, which have no source heading.
`part_id`	`VARCHAR`	The report part this block was parsed from, spelled as `committee_reports.part_id`, so (package_id, part_id) is the parent row. Appended last.