[Question] Why can't spaces be used when checking if a redis key exists?

[rhalstea]

When checking if a redis entry exists the key isn't allowed to have any spaces:
https://github.com/Azure/sonic-swss-common/blob/master/common/dbconnector.cpp#L609-L613

However, that restriction doesn't seem to be enforced when creating (i.e. SET) or reading keys (i.e. KEYS). I was able to create an entry like:

VRF_TABLE:"vrf 80"

Is there a technical reason for this? I guess we can use KEYS to check for existence, but the O(1) time complexity of exists is much nicer.

[prsunny]

I think even if you could create an entry with space in DB, the underlying infra may fail since some of these translates directly to create an entry in linux kernel where space is not allowed. So IMO, this is a general restriction.

[kcudnik]

in general is better to avoid spaces, in some places this can lead to regex wrong matching etc, in this case space it could be replaced by '=' or ':'

[rhalstea]

Granted VRF_TABLE was a contrived example and probably shouldn't have a space in the key. My question is more generic (e.g. maybe an ACL or some policy name). AFAICT today nothing is stopping a user from creating a key with a space:
https://github.com/Azure/sonic-swss-common/blob/master/common/dbconnector.cpp#L642-L647

The no space rule only seem to apply if the user wants to check if an entry already exists. Presumably the underlying infra would see an issue when the entry is created., Not when something checks if it already exists?