<< Back to previous view

[MATCH-89] perf 3x slower for ragged rows Created: 27/Oct/13  Updated: 28/Oct/13

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: kovas boguta Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

In both cases, match should fail immediately on seeing the first element.

Slow case:

(let [f
        (fn [x] (match x [1 1] 1 [2] 2 :else 3))]
    (time (dotimes [_ 100000] (f [3]))))
"Elapsed time: 57.899 msecs"

Fast case:

(let [f
        (fn [x] (match x [1 1] 1 [2 _] 2 :else 3))]
    (time (dotimes [_ 100000] (f [3]))))
"Elapsed time: 16.998 msecs"


 Comments   
Comment by David Nolen [ 27/Oct/13 8:24 PM ]

Ragged rows are hard to optimize. We used to try and intelligently group patterns - but this "optimization" created an incredible number of edgecase bugs. So until I or someone else can come up with a sound way to do this, going to continue to punt on this.

Comment by kovas boguta [ 28/Oct/13 6:03 PM ]

Can you point commit where that optimization existed? Just want to get a sense of what code path are affected.

Comment by David Nolen [ 28/Oct/13 9:08 PM ]

If you look at the repo tree around the first 0.2.0 alpha you'll see it, however the optimization was not global - it was a specific thing for each type of pattern. Again, I'm not particularly inclined to revisit this myself, but if you have a clever idea how it can be done without the previous problems I'm all for it.





[MATCH-87] Bad interaction with core.async Created: 11/Sep/13  Updated: 01/Dec/13

Status: Reopened
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Defect Priority: Minor
Reporter: Stefan Fehrenbach Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: bug
Environment:

[org.clojure/core.async "0.1.0-SNAPSHOT"] (more specifically core.async-0.1.0-20130827.050117-78)
[org.clojure/core.match "0.2.0-rc5"]



 Description   

This prints "foo" as expected:

(defn works []
  (.log js/console
        (match ["qux" "foo"]
          ["qux" "bar"] 1
          ["qux" x]     x)))

This fails with an exception: Uncaught Error: No matching clause: ["qux" "foo"]
in up-to-date Chromium

(defn fails []
  (go
   (.log js/console
         (match (<! (go ["qux" "foo"]))
           ["qux" "bar"] 1
           ["qux" x]     x))))

I don't know either core.async nor core.match very well. My best guess is a undesired interaction between the nonlocal control flow for backtracking and the core.async state machine.

Note that this is a version of core.async where http://dev.clojure.org/jira/browse/ASYNC-15 is fixed.



 Comments   
Comment by David Nolen [ 11/Sep/13 12:24 PM ]

This is not a bug, it's not possible to use <! outside of go blocks.

Comment by Stefan Fehrenbach [ 11/Sep/13 12:48 PM ]

Does that count as outside? Anyway, this should not be outside. Same problem.

(defn fails []
  (go (let [m (<! (go ["qux" "foo"]))]
        (.log js/console
              (match m
                ["qux" "bar"] 1
                ["qux" x] x)))))
Comment by David Nolen [ 11/Sep/13 1:17 PM ]

Sorry missed the surrounding go! However note tickets about interaction between core.match and core.async are extremely low priority without more information about whether the bug is actually in ASYNC or in MATCH. I just don't have the bandwidth to do that sort of investigation at the moment.

Still, thanks for the report.

Comment by Stefan Fehrenbach [ 11/Sep/13 2:05 PM ]

I see. I'm not so sure whether I should report this to ASYNC, any thoughts?
I'm not familiar with either library, but I'll try to investigate further when I have the time.
For now, I guess I'll do more stupid matching. It really is a shame, core.match is a perfect fit for dispatching on messages and channels coming from alts!. Anyways, thanks for your work!

Comment by David Nolen [ 30/Nov/13 2:28 PM ]

Is this still a problem? I think core.async recently added better analysis of the dot form.

Comment by Stefan Fehrenbach [ 01/Dec/13 2:45 AM ]

Hi, thanks for thinking about this. I got side tracked and did not look into it any further.
Unfortunately it still does not work.

Tested versions:
[org.clojure/clojurescript "0.0-2080"]
[org.clojure/core.async "0.1.242.0-44b1e3-alpha"]
[org.clojure/core.match "0.2.0"]





[MATCH-72] complexity analysis Created: 18/Jun/13  Updated: 18/Jun/13

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

For smaller matches avoid backtracking actually greatly improves performance at the potential cost of code size. Perhaps we can devise a heuristic such that we employ the the best option available? Would need some serious hammock time.






[MATCH-65] cata matching Created: 23/Nov/12  Updated: 15/Jun/13

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

Dan Friedman's pattern matcher has a nice feature called cata-matching - allowing recursive matching from the match itself. Useful when writing compilers.






[MATCH-64] Improve match compile times Created: 15/Aug/12  Updated: 15/Jun/13

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None





[MATCH-9] use case for dispatching on literals Created: 04/Sep/11  Updated: 10/Oct/11

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

Would speed up literal matching considerably.






[MATCH-8] Implement (p|q)ba heuristics Created: 04/Sep/11  Updated: 04/Sep/11

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

We should implement the rest of Maranget's suggested heuristics.

p - needed prefix
q - constructor prefix
b - branching factor
a - constructor arity






[MATCH-2] Matching Diagnostics Created: 04/Sep/11  Updated: 16/Jun/13

Status: Open
Project: core.match
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: David Nolen Assignee: David Nolen
Resolution: Unresolved Votes: 0
Labels: None


 Description   

Communicate to the user precisely what failed to match. Conversation here: http://groups.google.com/group/clojure/browse_thread/thread/675456fba1712214. We adopt the behavior of condp since that is closer to what match does and will do (predicate dispatch)



 Comments   
Comment by David Nolen [ 05/Sep/11 10:40 AM ]

Continuing the conversation from GitHub - concerning your changes, we should probably show what current occurrence failed to match as well as the breadcrumb.

Comment by David Nolen [ 26/Sep/11 7:10 AM ]

Matching diagnostics will be complicated by MATCH-1. Clojure doesn't yet support simple data conveying exceptions. Will probably have to use a combination of proxy and definterface.

Comment by Ambrose Bonnaire-Sergeant [ 30/Sep/11 4:39 AM ]

I haven't yet kept up with your backtracking changes, and I just noticed this conversation, so I'll have to have another look at diagnostics with backtracking.

Comment by David Nolen [ 16/Jun/13 1:57 PM ]

Lowering priority. Happy to take a patch, but I don't think I'll be looking into this myself in the near future.





Generated at Mon Sep 01 02:59:46 CDT 2014 using JIRA 4.4#649-r158309.