You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Python adapter: capture only for commands that can change the page #562
The JavaScript adapters take a per-action capture only for commands that can change the page; reads, waits, lookups and assertions reuse the last capture. The Python adapter captures on every wire command, reads included.
Measured in #516 (v11.2.0): on the Android Clock flow, 42 captures (21 findElements, 9 getElementText, 4 isElementEnabled, ...) where WebdriverIO takes 8 for the same flow; capture is 44% of wall time. On desktop Chrome, 26 of 42 captures were findElement, isDisplayed and getCurrentUrl.
Proposal
Apply the same rule in the Python adapter (mirror the JS classification, with a mirror-vectors.json section so both sides agree). It is lossless: those rows show the capture of the last command that took one, exactly as on the JS adapters. Expected saving about 5.7 s of a 22 s Android run.
Re-measure with DEVTOOLS_CAPTURE_TIMING=1; update the docs that describe which actions capture the page.
Activity