More optimizations

[alexander-yakushev]

Here are some more easy optimizations discovered while benchmarking Metabase's usage of HoneySQL.

1. clojure.string/split uses the regex machinery which is generally more wasteful than searching for a substring manually. HoneySQL so far only looks for . when splitting names. Besides, plenty of names don't have . in them and they still pay the regex tax. The new splitter function allocates nothing in such case.
2. Clojure's Keyword object already contains a corresponding Symbol, so we can return it in kw->sym (if there is no namespace). We could actually just call (symbol kw) for the same effect, but I think this behavior was introduced in Clojure 1.10, and HoneySQL claims to support 1.9.
3. I again tried to address the into's transient-persistent roundtrip. This time, I'm not making any speculative and potentially damaging changes. There are a lot of places in the code where into is used multiple times in succession, hence into* (name is up to be changed) both makes the code cleaner and prevents unnecessary transient roundtrips inbetween.

[alexander-yakushev]

Looks like (.sym kw) doesn't fly with Babashka. Perhaps, @borkdude may chime in if this is intentional?

[borkdude]

@alexander-yakushev You can add this:

(:bb (symbol (name k)))

before the :clj branch

[borkdude]

Btw since clojure 1.10 or so (symbol kwd) also just works I believe, so dropping clojure 1.9.0 support may also help

[alexander-yakushev]

Yes, I mentioned it in the PR. I'm not sure if this optimization is a good reason to drop 1.9 if there are no others.

[alexander-yakushev]

The custom :bb feature flag worked, thanks!

[seancorfield]

Given 1.9 usage has dropped to just over 2.5%, I am considering dropping support for it. I'll ask in a few places to see how badly that would inconvenience folks...