The @ prefix is an mpd convention: it is used in mpd.confbind_to_address and by libmpdclient (and everything downstream: mpc etc...). It doesn't hurt to be more permissive and accept paths starting with the NUL byte as well, as that's the Linux socket API convention and doesn't violate mpd2's backwards compatibility anyways.
Regarding backwards compatibility: for mpd.base, this should not be a breaking change: @ is not allowed in hostnames and connect(host) only accepts absolute paths. However, this may be a breaking change for mpd.asyncio users: the current implementation of connect(host) only checks for presence of the path separator / in host, meaning it accepts relative paths. I purposefully did not change this behavior in order to ensure the PR remains focused on one feature but this does mean that there is potential for backwards incompatibility when connecting to a path relative to cwd with a name starting with @. Of course, this is a pretty unlikely edge case, but I thought it'd be worth mentioning.
I did not notice any tests specifically testing connect() behavior in mpd/tests.py and therefore did not add my own. If I should add a test for this new feature, do let me know.
Mic92
commented
2 years ago
The
@prefix is an mpd convention: it is used inmpd.confbind_to_addressand by libmpdclient (and everything downstream: mpc etc...). It doesn't hurt to be more permissive and accept paths starting with the NUL byte as well, as that's the Linux socket API convention and doesn't violate mpd2's backwards compatibility anyways.Regarding backwards compatibility: for
mpd.base, this should not be a breaking change:@is not allowed in hostnames andconnect(host)only accepts absolute paths. However, this may be a breaking change formpd.asynciousers: the current implementation ofconnect(host)only checks for presence of the path separator/inhost, meaning it accepts relative paths. I purposefully did not change this behavior in order to ensure the PR remains focused on one feature but this does mean that there is potential for backwards incompatibility when connecting to a path relative to cwd with a name starting with@. Of course, this is a pretty unlikely edge case, but I thought it'd be worth mentioning.I did not notice any tests specifically testing
connect()behavior inmpd/tests.pyand therefore did not add my own. If I should add a test for this new feature, do let me know.