What I’m doing:

Reformatting dates ranging from early 1900’s until now, using Python’s datetime.strptime method. Straightforward.

What I saw upon inspection:

I find that I’ve lost all dates before 1969. Noooo!

What I thought happened:

Maybe all dates before 1969 were formatted in a way I hadn’t expected… no, they were the same… maybe all observations dated before 1969 didn’t comply with another part of my cleaning stage… no, they were real similar to all the others…

What actually happened:

Python’s datetime.strptime method added a century to every date before 1969. In other words, 1/1/68 became reformatted as 1/1/2068. In a later processing stage, I had removed all dates from the future, because my data is historical and future observations were impossible.

What the docs say:

“When 2-digit years are accepted, they are converted according to the POSIX or X/Open standard: values 69-99 are mapped to 1969-1999, and values 0–68 are mapped to 2000–2068.”

Takeaway:

Sometimes (always) it’s a good idea to reference the documentation of modules when code isn’t behaving as expected.

Also:

Dates are a pain.