path_to_parts parsing bug about sanic-routing

[ahopkins]

Hmm, I think it might be worth it to scrap the regex splitting in this function. I have tested it quite a bit, but it does seem like this is a use case that breaks it. Perhaps a solution will be to run a regular .split("/") operation, and then loop through the segments looking for instances where there is an open < but the close > is in the next segment, and merge.

[ahopkins]

def _normalize(parts, delimiter):
    part_iter = iter(parts)
    normalized = []
    for part in part_iter:
        if part.startswith("<") and not part.endswith(">"):
            ending = ""
            try:
                while not ending.endswith(">"):
                    ending += delimiter + next(part_iter)
            except StopIteration:
                raise Exception("invalid")

            part += ending
        normalized.append(part)
    return normalized


def path_to_parts(path, delimiter="/"):
    parts = _normalize(path.split(delimiter), delimiter)
    return tuple([part if part.startswith("<") else quote(part) for part in parts])

I just threw together this little snippet to prove my point. I think this is probably the way to go and should be more accurate in catching corner cases like this one.

---

/image/iiif/<image_id>/<region:full|square|\d+,\d+,\d+,\d+>/<size:max|\d+,|,\d+|\d+,\d+>/<rotation:int>/default.jpg
('', 'image', 'iiif', '<image_id>', '<region:full|square|\\d+,\\d+,\\d+,\\d+>', '<size:max|\\d+,|,\\d+|\\d+,\\d+>', '<rotation:int>', 'default.jpg')

[kserhii]

Can you maybe provide a path for which normal .split("/") does not work? Just to have a clear understanding of the problem.

[ahopkins]

Can you maybe provide a path for which normal .split("/") does not work? Just to have a clear understanding of the problem.

@app.get(r"/path/to/<deeply_nested:[a-z/]+>/thing")
async def handler(request, deeply_nested):
    return text(deeply_nested)

$ curl localhost:9999/path/to/location/with/slashes/thing
location/with/slashes

---

There are three use cases:

1. Custom regex routes that have a / in them
2. a <path:path> param (which uses item 1)
3. static files (which uses item 2)

---

Given this, that is the reason for the complication with my proposed solution. Using .split() will unintentionally break apart some of these segments, so we will need to stitch them back. I thought the regex covered all scenarios, but I think a "dumber" solution might just be simpler.

[kserhii]

I've looked a bit and found the solution her:
https://stackoverflow.com/questions/732029/how-to-split-string-by-unless-is-within-brackets-using-regex

Here is the pull request with extra tests

[ahopkins]

Fixed by #40