Quamina Priorities

[timbray]

For a long time I've had the states-explosion problem at the front of my mind and now that it's solved, I can start to think of other things to do. Some of the most obvious forward directions are:

More optimization

Q's performance is already good, except for the corner case with wildcards where, if you combine a few thousand, you can slow Q down by a couple orders of magnitude. Fixing this is probably the top optimization prioritization item. There's no guarantee of a fix (AWS-Event-Ruler still suffers from a flavor of this problem) but I have some strong ideas, one of them being turning the NFA into a DFA.

Implement Ruler-compatible wildcard patterns

These are the same as our shellstyle patterns but have escaping, so you can say:

{"x": [{"wildcard":"Hey \\*Joe\\*, get the *"}]}`

So, the task is to generalize \-escaping.

Implement a bunch of Ruler-compatibility changes

There are a bunch of Issues open for this. These should be fairly straightforward with the newer, cleaner, NFA-building approach. If someone feels like they'd like to do feature-building, this is a strong candidate.

One that's special is the general-purpose numeric value handling. There's code sketched in but never actually turned on.

Implement general regular expressions

This will be more intellectually challenging but probably intense fun. It also has the advantage that if you do this, probably many of the Ruler-compatibility things can be expressed as a regular expression.

What do you think?

I have opinions but would like to hear others.

[embano1]

Regarding

These are the same as our shellstyle patterns but have escaping

Basically you're saying be 100% aws/ruler compatible to allow:

Wildcard matches only work on string-valued fields. A single value can contain zero to many wildcard characters, but consecutive wildcard characters are not allowed. To match the asterisk character specifically, a wildcard character can be escaped with a backslash. Two consecutive backslashes (i.e. a backslash escaped with a backslash) represents the actual backslash character. A backslash escaping any character other than asterisk or backslash is not allowed.

[timbray]

Exactly. The idea is, once Q has compatible wildcard matching we'll deprecate (but never remove) shellstyle.

For the same reason, I have another PR coming to do the first steps of numeric matching. Q already had number.go (not yet used in practice) which is similar to Ruler's ComparableNumber but with 2 digits more precision. But I'm going to cut it back to have the exact same semantics as Ruler, because I think compatibility is a big deal.

[embano1]

Got it!