BUG(pydarshan): `id` field in `to_df()` output should be `uint64` instead of `int64`

[jay-tau]

TL;DR: The id field in the DataFrame generated by DarshanRecordCollection.to_df() should be of type uint64 instead of int64.

Currently, the Pandas DataFrames created from PyDarshan POSIX records (via report.records[modulename].to_df()) represent the id field as a signed 64-bit integer (int64). However, this field originates from the Darshan C library, which uses a hash function (darshan_hash()) that returns values of type unsigned long long—corresponding to a 64-bit unsigned integer (uint64).

To maintain fidelity with the underlying data representation and avoid potential issues with signed integer overflow or interpretation, it would be more appropriate to cast the id field to uint64 in the resulting DataFrame.

References:

Implementation of darshan_hash: darshan-util/darshan-common.c

Header for darshan_hash: darshan-util/darshan-common.h

Slack: Issue, Response from @shanedsnyder