issue_comments

I may have to do extra work here python def database(self, database, format=None): db = self.ds.databases[database] if self.ds.setting("hash_urls") and db.hash: path = self.path( f"{dash_encode(database)}-{db.hash[:HASH_LENGTH]}", format=format ) else: path = self.path(dash_encode(database), format=format) return path The URLs that incorporate a hash have a dbname-hash format - will that - in the middle there mess up the dash decoding mechanism? I think it will.

Might be able to solve that like so:

python async def resolve_db_name(self, request, db_name, **kwargs): hash = None name = None decoded_name = dash_decode(db_name) if decoded_name not in self.ds.databases and "-" in db_name: # No matching DB found, maybe it's a name-hash? name_bit, hash_bit = db_name.rsplit("-", 1) if dash_decode(name_bit) not in self.ds.databases: raise NotFound(f"Database not found: {name}") else: name = dash_decode(name_bit) hash = hash_bit else: name = decoded_name

• Maybe use dash encoding for database name too?

Yes, I'm going to do this. At the moment if a DB file is called fixx%tures.db when you run it in Datasette the path is /fix%2525tures - which is liable to break.

Just spotted this: https://github.com/simonw/datasette/blob/de810f49cc57a4f88e4a1553d26c579253ce4531/datasette/views/base.py#L203-L216

Maybe the db name should use dash encoding too?

If so, relevant code includes this bit:

https://github.com/simonw/datasette/blob/de810f49cc57a4f88e4a1553d26c579253ce4531/datasette/url_builder.py#L30-L38

For consistency, I'm going to change how ?_next= tokens work too. Right now they work like this:

https://github.com/simonw/datasette/blob/de810f49cc57a4f88e4a1553d26c579253ce4531/datasette/views/table.py#L501-L507

https://github.com/simonw/datasette/blob/de810f49cc57a4f88e4a1553d26c579253ce4531/datasette/utils/init.py#L114-L116

I'm going to change those to use dash-encoding instead.

I considered looking for % in those values and replacing that as - too, but since Datasette isn't 1.0 yet I'm going to risk breaking any pagination tokens that people might have saved away somewhere!

Updated documentation:

Change of plan: based on extensive conversations on Twitter - see https://github.com/simonw/datasette/issues/1439#issuecomment-1059851259 - I'm going to try a variant of this which is basically percent-encoding but with a hyphen instead of a percent symbol.

Reason being that the current scheme doesn't handle the case of % being part of the table name, which could cause weird breakage due to some proxies decoding percent encoding before it gets to Datasette.