arpitjain099 commented on PR #68454:
URL: https://github.com/apache/doris/pull/68454#issuecomment-5811482761
Thanks, pushed a follow-up.
On the tests: the escape matrix now covers the one-hex-digit forms (`%z1`,
`%1z`, `a%1gb`), the sign and space cases a stream parse is most willing to
swallow (`%+1`, `%-1`, `% 1`, `%1 `), a `%` at and one character from the end,
both digit cases, `%00`, multi byte UTF-8, and an encode/decode round trip over
every byte value from 1 to 255.
The review is right that validating the path name was not enough, and the
reason turned out to sit outside this file. `init_from_evhttp` parses the query
string into `_params`, then `_find_handler` routes and `PathTrie::put` adds the
captures with `map::insert`, which does not overwrite. So the query string won:
```
PUT /api/realdb/tbl/_stream_load?db=otherdb -> param("db") = otherdb
```
`put` now assigns, and the same request gives `realdb`. There is a case for
it in `path_trie_test`.
I also clear `ctx->db` and `ctx->table` when the decode fails. `url_decode`
clears its output and then appends until it gives up, so the log line and the
failed load record were both showing a truncated name.
One point in the review I do not think holds: that FE folding and BE runtime
still disagree on signed escape forms such as `%+1`. I ran both over the same
19 inputs and they agree on all of them now, `%+1` included. Before this change
the BE decoded it to `\x01` while `URLDecoder` rejected it, which is the
divergence the patch removes.
I cannot run the BE suite on this machine, so CI is the real check on all of
the above. Separately, `path_trie_test.cpp` does not compile on its own for me
because `SplitTest` calls the private `split`; that looks unrelated but might
be worth a glance.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
