If your regex matches a single quote(') anywhere, the Elisp parser for Emacs would throw an error because the surrounding quotes would be double quotes and not single quotes.

Currently, the point is moved to the beginning of the match. It would be nice to highlight the match to make it easier to see the match.

Currently, all tests only test for re.MULTILINE as re.MULTILINE is the default for evil_search.py. However, this is not the default for the function.

When you run s/^hello/bye on the following:
*hello
hello
hello

where * is the cursor position... We get:
*hello
bye
bye

Expected behavior:
*bye
bye
bye

Benchmark current implementation against the builtin implementation to see how much of a performance degradation we have. Also, profile each major step of the current implementation (ie sending data to Python, parsing it on the Python backend, collecting the data, parsing the data from Python, searching and replacing the text).

Current set of test only covers a small portion of the Python backend. This makes it harder to debug and verify that changes did not break anything.

The Emacs file has the basic framework for a Emacs package, but still missing the meat of it. Possibly try to add this to Melpa afterward.

When you search on Emacs splash screen, the cursor sometimes jump to the wrong place (possibly due to how links are treated).

Given the regex, s/he|el/bye, on the string 'hello', it produces: (['he', '0', 'el', '1'], '').

Expected: (['he', '0'], '')

Current test doesn't cover the cases when the search results overlap.

Unicode support needs more testing. Probably need to modify the base64 encoding/decoding system to include Unicode support.