Repository navigation
refactor Config() and _ENV_CONFIG聽#3028
Description
Activity
Should
_ENV_CONFIGeven exist? its only used inload_catalogandlist_catalogsIf we want
Config()to be cached, maybe it should be in config.py
We should still make it lazy and not loaded automatically on importWe might want to replace all calls to
Config()in the codebase with the cached version, so that we wont need to re-parse everytimeReacted by Drew and jj.leeI reviewed the agreed direction in #3083 and the follow-up #3208. Both PRs were closed automatically as stale and there is no open successor, so I prepared a narrow refreshed branch for the first slice: keep the existing
Config()API, defer configuration-file and environment reads until the firstconfig/get_*access, and cache the merged result per instance. It does not move_ENV_CONFIGor change the catalog-loading API.The branch is https://github.com/emecii/iceberg-python/tree/refactor/lazy-config-loading at
5586c286. On currentmain, the full unit target passes (4,019 passed, 3 skipped), as domake lintand Apache RAT on a clean worktree.@jx2lee @kevinjqliu, is it okay for me to open this narrow PR, or would you prefer the earlier work to be revived/continued instead?
OpenAI Codex generated the patch, tests, and this comment and ran the reported checks. No human review is asserted.
Apache Iceberg version
None
Please describe the bug 馃悶
Follow up to #3006, i want to see if we can refactor Config() and _ENV_CONFIG
a couple of issues ive seen
__init__reads both env vars and local config files implicitly, this can be a potentially footgun.#3006 is a fix for pytests. Ideally, I think we can be more explicit here
Willingness to contribute