asOf=YYYY-MM-DD on GET /us/statutes/section/{actId}/body returns the section’s text as it stood
on that date. It costs the same 6 credits as an ordinary body fetch.
This matters when the question is dated: what an employer’s obligation was on the day of the
conduct, which version of a rule a contract signed in 2019 incorporated, what a regulation required
before an amendment. Answering it from today’s text is how a confidently wrong answer gets built.
Every answer says where it came from
The response carries anasOf block. Read source before you use the text, because the same
parameter is answered by two different mechanisms and one of them is an estimate.
A fourth outcome is not a
source value at all: existed: false means the earliest thing we ever
captured for this section is its addition, and that addition postdates your date, so it
demonstrably did not exist yet. That is an answer, arguably the most valuable one this parameter
gives, so it is charged.
Pass
basisChangeId to
GET /us/statutes/section/{actId}/changes to see the
event the answer rests on.
An
asOf response never populates html, on any corpus. Reconstructed text is plain text, and
synthesising markup the publisher never printed would defeat the point of the endpoint. Use
text, or content where the corpus supports it.Reconstruction versus a published edition
asOf resolves two different ways depending on the corpus, and the difference decides how much
weight the answer carries.
A published edition
For corpora where the publisher issues dated editions,
asOf resolves to the edition that
was in force on your date, served as published. This is the publisher’s own artifact, not our
estimate.A reconstruction
Everywhere else, the corpus holds one current text per citation and the earlier one is rebuilt
from observed change history. Accurate to what we captured, bounded by when capture began.
Corpora that answer from a published edition
Two behaviours to expect on
CFR_ANNUAL. editionsObserved is an enumeration, never a range:
a version observed in 1998, 1999 and 2003 lists exactly those years, and a date in an unobserved
gap returns asOf.isBounded: false rather than the nearest year held. A citation year is the
volume’s own <REVISED> year, not the folder GPO filed it under, because GPO reprints unchanged
volumes forward without re-dating them.
Where history is a separate document rather than a date
Some corpora publish their own prior versions, so you retrieve history by citation instead of by date.whd_opinion_letter keeps withdrawn and superseded letters retrievable under their own
act_id, including the 2009 letters WHD re-issued in 2018 under new numbers. A re-issued pair
shares one PDF at the publisher but is two documents with opposite status.
Where point-in-time is weakest
Corpora whose publisher replaces text in place and issues no superseded editions, such as the CMS Medicare manuals and the DOL Field Operations Handbook, give us nothing to reconstruct from beyond our own capture history. ExpectisBounded: true on dates before capture began.
Worked example
isBounded: true here because 2019 predates observedFrom. The text is our best reconstruction,
and a product built on it should say so to its reader.
Choosing the right tool
Status and currency
actStatus, goodLawStatus, and whether a section is still operative.Federal coverage
Which corpora hold editions, and how far back each one reaches.