[godhiraj-code]

What this fixes

After PR #9, every wrapped find_element() also called find_elements() with the same locator and passed every match into the fingerprint script to record the element's index. That cost grew with the number of matches (about 0.37 to 0.41 s per lookup for a selector matching 2000 elements in my local measurement on 77e0de7). The index never added identity information: find_element() and the stale retry both take the first match, so it was always 0.

Changes

• Fingerprint is now tag, visible text, id, name, data-testid, aria-label, and the parent's visible text (both capped at 200 characters), read in one script call. No scan of all locator matches during find_element() or the stale retry.
• Elements from find_elements() or scripts no longer read a fingerprint on successful clicks. Only when the enabled wait times out does one diagnostic script read add the tag and text to the error.
• README: documents that a change in the parent's text (for example Cart (1) to Cart (2)) makes the stale retry refuse, and that a specific locator does not bypass that check.
• CHANGELOG 1.1.0 entries updated to match.
• Tests: unit and browser regressions for the removed scan, the lazy diagnostics and the parent text refusal.

Not done

Not merged, tagged, released or published. 1.1.0 is still unreleased.