Catalog connector

Written after the fact. The connector was declared and wired in 09ae433, and no decision record named it until “What approves a connector” found four connectors in that state. It is the least uncovered of the four: “The catalog of budget line items” scoped the source in detail, probed it, and deliberately did not start the extraction. What that record does not do is decide whether the Catalog is a connector at all or a second source under lsc-budget, because in the phase it was written there was nothing to attach either way.

The extraction phase then had to decide it, and did so in a commit message. Eighteen editions of the education volume, 12,048 money rows and 2,008 line items, and the series the corpus had wanted since genesis: GRF 200550 Foundation Funding, enacted, one figure per fiscal year from FY2006 to FY2027 with no year missing and none claimed by two editions.

Declare lsc-catalog as a connector in its own right rather than as sources under lsc-budget, Wired with a recorded remainder, feeding fiscal-period, program, legislation and parameter.

Same publisher, different artifact — the reason ohio-session-laws is not folded into ohio-laws and tax-casino is not folded into tax-abstract. lsc-budget retrieves the greenbook, which is LSC’s analysis of a specific act as enrolled. The Catalog is LSC’s standing reference, restating each line item’s history as of publication. A figure restated four editions later is not the same claim as a figure published at enactment, and the corpus’s appropriation nodes cite the latter. One connector holding both would have one status and one blocker for two claims that a reader must be able to tell apart.

Keep the vintage. Every row carries the edition it came from, which is what makes the two claims distinguishable at all rather than merely distinguished in prose here.

Pair columns by count, not by position. The greenbook workbook attempt failed reading nearest header label, because labels are narrower than the figures beneath them. Here the years, the labels and the amounts agree in number on all 566 sampled blocks once --- is read as the stated absence it is, so order is the pairing and no position arithmetic is involved. A block whose three counts disagree is refused and named rather than guessed at.

Both bienniums the greenbook route cannot reach are now held. FY2006-07, whose greenbook has no line-item table at all, is in the 2006 edition. FY2012-13, whose two workbook variants LSC serves as one file, is in the 2011 edition.

The defect that reverted the workbook attempt is labeled here rather than silent. An edition published at a biennium’s start carries two Appropriation columns; the next carries one Adj. Approp. in their place — the enacted figure revised by what has happened since. That is the same supersession that produced zero appropriation rows for a whole biennium without failing. The difference is that the publisher names it, so the only way to repeat that failure is to throw the label away. Three kinds, never two, and a test asserts no edition mixes them.

Eighteen editions, not the nineteen the scoping record claimed. The 2012 URL resolves and returns a PDF; it is the 2011 document byte for byte — same SHA-256, same length. That record verified existence by status code and content type, and only a digest catches a file served under two names. Wiring both would have put every FY2012 and FY2013 enacted appropriation in the fixture twice under two vintages that are one document. The scoping record is amended in place and the test holds the count.

Four fiscal-period nodes stop resting on an index page. They cited one for their appropriating act with an explicit caveat that the Am. Sub. designation came from a greenbook not held here. The Catalog confirms it for H.B. 59 and H.B. 64 — and prints H.B. 49, H.B. 166, H.B. 110 and H.B. 33 bare on every one of hundreds of mentions. That is chronological, so it reads as LSC’s house style changing at the 132nd rather than as a fact about four bills, and it is recorded as unresolved rather than corrected in either direction.

The remainder is what the Catalog is, not what it is missing. Wired for the education volume of every edition that has one — 2006 and 2008 through 2025. 2007 has no edition. And the earliest edition reaches FY2002, so the Catalog thickens the pre-2000 record rather than extending it: the Foundation Program era and DeRolph I still need the session laws.

Add the editions as sources under lsc-budget. Rejected, above. It is the economical move and it costs the distinction between an enacted figure and a restated one, which is the single thing this source is uniquely good for.

Take the whole-catalog PDF rather than the education volume. Rejected in the scoping record and the reasoning held: the education volume is the unit this repository reasons about, the per-agency split is the publisher’s own, and pinning a digest on a document that changes whenever any of a hundred agencies changes would make every unrelated edit look like a source revision.

Treat the Catalog as a replacement for the greenbook and session-law work. Rejected in the scoping record, and the extraction confirmed why: the Catalog’s earliest edition reaches FY2002 and the greenbook’s earliest actuals are FY1999, so it does not even cover the range it would have had to replace.

Wire the 2012 URL as a nineteenth edition, as the scoping record’s count implied. Rejected on the digest. Keeping it would have been the same pathology the-greenbook-series found in the workbooks, caught this time before it reached a fixture rather than after.