type-inference on namespace constants

[Michael Blume]

If I create a local with (let [cs (Charset/forName "UTF-8")] ...) the local cs is automatically tagged with the type Charset, and any java method calls within the block will be inferred using that information.

If I create a namespace global with (def cs (Charset/forName "UTF-8")) the var cs is *not* tagged with the type Charset. If I want to prevent reflection in calls that use it, I need to write (def ^Charset cs (Charset/forName "UTF-8")).

I have seen a lot of reflection in various libraries that went away when top-level constants were type-hinted, and which wouldn't have shown up in the first place if their types were inferred. Is there a reason for the current behavior? Should it be changed?

[Andy Fingerhut]

Top-level constants are hard for the compiler to determine are truly constants, and not alter-var-root'ed or binding'd elsewhere?

Andy

--
You received this message because you are subscribed to the Google Groups "Clojure Dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to clojure-dev...@googlegroups.com.
To post to this group, send email to cloju...@googlegroups.com.
Visit this group at http://groups.google.com/group/clojure-dev.
For more options, visit https://groups.google.com/d/optout.

[Nicola Mometto]

It's a tradeoff to allow easier REPL driven development, where you can redefine a Var on the fly.
Also to account for alter-var-root