Skip to content

bbh_michigan_monthly.py docstring understates its network transfer (~310 KiB claimed, ~825 KB measured) #97

Description

@mmcky

Found during the independent validation run for #96 (section 3). Both builders' HTTP traffic was instrumented with a urllib wrapper logging every request's Range header, status, and bytes.

builders/bbh_macro_quarterly.py matches its docstring exactly: 4 requests, all Range/206, 296,475 bytes total, only zenodo.org contacted.

builders/bbh_michigan_monthly.py is also genuinely range-based — no full-archive fallback executed, only zenodo.org contacted — but it transfers substantially more than its docstring's "~310 KiB": 1 API GET (4,181 B) plus 6 range requests totalling 820,763 B (each of the three members is read through a 256 KiB buffer), for a grand total of 824,944 bytes. That is ~2.6× the stated figure, though still ~0.4% of the 198.8 MB archive.

Documentation-only change: update the docstring's transfer figure, or shrink the read buffer if the smaller number is meant to be a property of the builder. The range-reading claim itself, and byte-for-byte reproduction, both verified (see the #96 validation report).

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions