bytes hash to the record's source_sha256 and the .pdf.source sidecar names the record's source_url
located: a verbatim source line of this row appears on page 1 of the source PDF and on no other page the record read
Page 1, whole — no highlight
none of this row's 1 source lines occurs exactly once on page 1 (1 occur more than once or not at all in the word layer). Drawing a box would mean picking an occurrence, which is the defect this rule exists to stop; showing the whole page instead.
Page 1 of in_regs.pdf
The page is the document's, unedited. What this compilation read off it is the row above; what it did not read is on the state page under “what this record does not carry”.
