openvax / datacache / 36370375516

96%

master: 93%

Build Type

Pull #76

github

Committed by iskandr

Commit Message

Add get_cache_root for a shared cache root; import pandas lazily (1.12.0)

OpenVax packages read OPENVAX_DATA_CACHE as the root itself, else use the
platform cache directory named openvax, with a package variable such as
OSTEOSARC_CACHE first. get_data_dir appends its subdir to an environment
value, and pyensembl relies on that, so osteosarc and Vaxrank each
re-implemented the rule. Add get_cache_root(name, *envkeys): the first
variable set to a non-blank value is the root, trimmed and ~-expanded,
else user_cache_dir(name). Document the shared OpenVax convention and its
<root>/objects/sha256 layout in the integration guide.

Importing datacache also imported pandas, numpy, and requests, about
350 ms, which is why osteosarc called appdirs directly instead. Import them
only in the functions that use them; importing datacache now takes about
35 ms. download.pd and download.requests still resolve to those modules,
through a module __getattr__.

Also release the db_type mapping of numpy fixed-width types.

Claude-Session: https://claude.ai/code/session_01WCxJuZirP1wNSKkNYxZsE5

Pull Request Pull Request #76: Release 1.12.0: get_cache_root for a shared cache root, fast import

Coverage Stats

43 of 43 new or added lines in 5 files covered. (100.0%)

1 existing line in 1 file now uncovered.

1113 of 1157 relevant lines covered (96.2%)

7.68 hits per line