Firstly, I think you over estimate the cost of an if. Ifs are very very
fast, especially if you are doing identical? checks (like a case is doing).
A simple pointer comparison such as (keyword-identical? :foo :foo) is going
to be way faster than a hashmap lookup.
Secondly, ifs are very JIT friendly. Simpler code often means simpler
inlining. This could mean that several branches of the if/switch could be
removed if the JIT can prove that certain branches cannot be reached. This
is much harder to do with a hash map where things could change from one
dispatch to the next, and there isn't really a way to communicate
immutability to the JVM/JS JITs.
On the JVM I've done many tests while building core.async to try to find a
better way to do SSA block dispatching. In the end switch (aka case) won
every single time. A close second is a small set of nested ifs. Function
pointers are just expensive compared ifs and jump tables.
Timothy
On Tue, Feb 18, 2014 at 5:02 PM, t x <txrev...@gmail.com> wrote:
> With apologies for potential stupidity:
>
> * (get {... } key) is an operation that is O(log n) worst case (if
> using Red-Black trees) or O(1) average case (hashing)
>
> * (get { ... } key) does not have integer/string limitation
>
> Therefore: why are we not "compiling down to get" ? Are the big-Oh
> constants enormous?
>
> On Tue, Feb 18, 2014 at 3:51 PM, Michał Marczyk
> <michal.marc...@gmail.com> wrote:
> > Great! First cut at compiling case to switch (when given numbers and
> > strings only as tests at the moment, but this can be improved) here:
> >
> >
> https://github.com/michalmarczyk/clojurescript/tree/713-compile-case-to-switch
> >
> > I'll post more details on the ticket.
> >
> > Cheers,
> > Michał
> >
> >
> > On 18 February 2014 22:38, David Nolen <dnolen.li...@gmail.com> wrote:
> >> Due to asm.js some JS engines are starting to compile switch statements
> >> where the cases are integers into jump tables. Compiler enhancement that
> >> compiles optimizable case expressions to JS switch statements would be a
> >> welcome enhancement - http://dev.clojure.org/jira/browse/CLJS-713.
> >>
> >> David
> >>
> >>
> >> On Tue, Feb 18, 2014 at 4:35 PM, t x <txrev...@gmail.com> wrote:
> >>>
> >>> Looking at
> >>>
> https://github.com/clojure/clojurescript/blob/r2156/src/clj/cljs/core.clj#L1144-L1147
> >>> , you win. :-)
> >>>
> >>> On Tue, Feb 18, 2014 at 1:33 PM, Michał Marczyk
> >>> <michal.marc...@gmail.com> wrote:
> >>> > On 18 February 2014 22:31, Michał Marczyk <michal.marc...@gmail.com>
> >>> > wrote:
> >>> >> Actually in ClojureScript case dispatch is O(n), since it's
> >>> >> implemented as a macro expanding to cond.
> >>> >
> >>> > As of release 2156:
> >>> >
> >>> >
> >>> >
> https://github.com/clojure/clojurescript/blob/r2156/src/clj/cljs/core.clj#L1119
> >>> >
> >>> > --
> >>> > You received this message because you are subscribed to the Google
> >>> > Groups "Clojure" group.
> >>> > To post to this group, send email to clojure@googlegroups.com
> >>> > Note that posts from new members are moderated - please be patient
> with
> >>> > your first post.
> >>> > To unsubscribe from this group, send email to
> >>> > clojure+unsubscr...@googlegroups.com
> >>> > For more options, visit this group at
> >>> > http://groups.google.com/group/clojure?hl=en
> >>> > ---
> >>> > You received this message because you are subscribed to the Google
> >>> > Groups "Clojure" group.
> >>> > To unsubscribe from this group and stop receiving emails from it,
> send
> >>> > an email to clojure+unsubscr...@googlegroups.com.
> >>> > For more options, visit https://groups.google.com/groups/opt_out.
> >>>
> >>> --
> >>> You received this message because you are subscribed to the Google
> >>> Groups "Clojure" group.
> >>> To post to this group, send email to clojure@googlegroups.com
> >>> Note that posts from new members are moderated - please be patient with
> >>> your first post.
> >>> To unsubscribe from this group, send email to
> >>> clojure+unsubscr...@googlegroups.com
> >>> For more options, visit this group at
> >>> http://groups.google.com/group/clojure?hl=en
> >>> ---
> >>> You received this message because you are subscribed to the Google
> Groups
> >>> "Clojure" group.
> >>> To unsubscribe from this group and stop receiving emails from it, send
> an
> >>> email to clojure+unsubscr...@googlegroups.com.
> >>> For more options, visit https://groups.google.com/groups/opt_out.
> >>
> >>
> >> --
> >> You received this message because you are subscribed to the Google
> >> Groups "Clojure" group.
> >> To post to this group, send email to clojure@googlegroups.com
> >> Note that posts from new members are moderated - please be patient with
> your
> >> first post.
> >> To unsubscribe from this group, send email to
> >> clojure+unsubscr...@googlegroups.com
> >> For more options, visit this group at
> >> http://groups.google.com/group/clojure?hl=en
> >> ---
> >> You received this message because you are subscribed to the Google
> Groups
> >> "Clojure" group.
> >> To unsubscribe from this group and stop receiving emails from it, send
> an
> >> email to clojure+unsubscr...@googlegroups.com.
> >> For more options, visit https://groups.google.com/groups/opt_out.
> >
> > --
> > You received this message because you are subscribed to the Google
> > Groups "Clojure" group.
> > To post to this group, send email to clojure@googlegroups.com
> > Note that posts from new members are moderated - please be patient with
> your first post.
> > To unsubscribe from this group, send email to
> > clojure+unsubscr...@googlegroups.com
> > For more options, visit this group at
> > http://groups.google.com/group/clojure?hl=en
> > ---
> > You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> > To unsubscribe from this group and stop receiving emails from it, send
> an email to clojure+unsubscr...@googlegroups.com.
> > For more options, visit https://groups.google.com/groups/opt_out.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clojure@googlegroups.com
> Note that posts from new members are moderated - please be patient with
> your first post.
> To unsubscribe from this group, send email to
> clojure+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "Clojure" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to clojure+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
--
"One of the main causes of the fall of the Roman Empire was that-lacking
zero-they had no way to indicate successful termination of their C
programs."
(Robert Firth)
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
---
You received this message because you are subscribed to the Google Groups
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
