DOC: Cleaned references to pandas <v0.12 in docs by topper-123 · Pull Request #17375 · pandas-dev/pandas (original) (raw)

I like this. I think the cutoff you've chosen here is a good one for now. v0.12 marks a slow-down in the development pace of pandas (just look at the pace of release tags).

0.17.0 may be >3 versions old, but it's also less than two years old. In general I would go by time instead of number of major versions.

In gotschas.rst, there is a sentence "As of pandas 0.11, pandas is not 100% thread safe." I haven't altered this, but I presume this still is correct in the newest version of pandas? Then IMO it should be changed to reference a newer version or simply to "pandas is currently not 100% thread safe."

I don't understand what this was originally intended to convey. There are at least two types of thread-safety:

1. Thread-safe libraries don't crash when run from multiple threads.
2. Thread-safe data structures have locks to ensure atomic operations on mutable data structures.

Like most Python code, pandas falls in the first category, not the second. This is tested routinely by dask. But not even built-in data structures are thread-safe in the second sense in Python.

In io.rst there is a sentence "0.10.1 of HDFStore can read tables created in a prior version of pandas, ...". I'm not even sure the "0.10.1" references the version of pandas or a HDF library and I left it alone. The paragraph also discusses backwards compatability, which makes it somewhat relevant to keep around, even if it's an old change.

I'm pretty sure 0.10.1 references the pandas version. HDF5 is on version 1.8.17.

I don't think it's important to mention the version here at this point, since 0.10.1 is quite old.