<< Back to previous view

[CLJ-1416] Support transients in gvec Created: 06/May/14  Updated: 05/Jul/14

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: Michał Marczyk Assignee: Michał Marczyk
Resolution: Unresolved Votes: 0
Labels: collections, transient

Attachments: Text File 0001-CLJ-1416-transients-hash-caching-for-gvec-Object-met.patch     Text File 0002-CLJ-1416-transients-hash-caching-interop-improvement.patch    
Patch: Code
Approval: Triaged

 Description   

Vectors of primitives produced by vector-of do not support transients.

core.rrb-vector implements transient support for vectors of primitives. Such transient-enabled vectors of primitives can be obtained in a number of ways: (1) using a gvec instance as an argument to fv/catvec (if RRB concatenation happens, which is not guaranteed) or fv/subvec; (2) passing a gvec instance to fv/vec, which as of core.rrb-vector 0.0.11 will simply rewrap the gvec tree in an RRB wrapper; (3) using fv/vector-of instead of clojure.core/vector-of. Native support in gvec would still be useful as part of an effort to make supported functionality consistent across vector flavours (see CLJ-787 in this connection); gvec is also simpler and still has (and is likely to maintain) a performance edge.

A port of core.rrb-vector's transient support to gvec is available here:

https://github.com/michalmarczyk/clojure/tree/transient-gvec

I'll bring it up to date with current master shortly.

See the clojure-dev thread for some benchmarks:

https://groups.google.com/d/msg/clojure-dev/9ozYI1e5SCM/BAIazVOkUmcJ



 Comments   
Comment by Michał Marczyk [ 13/May/14 5:32 AM ]

Here's the current version of the patch (0001-CLJ-1416-transients-hash-caching-for-gvec-Object-met.patch). It includes a few additional changes – here's the commit message:

CLJ-1416: transients, hash caching for gvec, Object methods for gvec seqs

  • Adds transient support to gvec
  • Adds hash{eq,Code} caching to gvec and gvec seqs
  • Implements hashCode, equals, toString for gvec seqs

https://github.com/michalmarczyk/clojure/tree/transient-gvec-1.6

Comment by Michał Marczyk [ 05/Jul/14 2:59 AM ]

Here's an updated patch with some additional interop-related improvements.

The new commit message:

CLJ-1416: transients, hash caching, interop improvements for gvec

  • Adds transient support to gvec
  • Adds hash{eq,Code} caching to gvec and gvec seqs
  • Implements hashCode, equals, toString for gvec seqs
  • Correctly implements iterator-related methods for gvec and gvec seqs
  • Introduces throw-unsupported and caching-hash (both marked private)




[CLJ-1386] Add transient? predicate Created: 17/Mar/14  Updated: 23/Mar/14

Status: Open
Project: Clojure
Component/s: None
Affects Version/s: Release 1.6
Fix Version/s: None

Type: Enhancement Priority: Minor
Reporter: Devin Walters Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: collections, transient
Environment:

N/A


Attachments: Text File 0001-Add-transient-predicate.patch     Text File 0002-Add-transient-predicate.patch    
Patch: Code and Test
Approval: Triaged

 Description   

I've encountered situations where I wanted to check whether something was transient in order to know whether I should call assoc! or assoc, conj! or conj, etc.

This patch adds `transient?` as a predicate fn.



 Comments   
Comment by Alex Miller [ 17/Mar/14 10:21 AM ]

Patch needs a docstring and a test.

Comment by Devin Walters [ 17/Mar/14 4:42 PM ]

Alex: I figured that would be the case! Sorry about that. I've updated the patch. It now includes a docstring and has tests of `transient?` for #{}, [], and {}.

Thanks!

Comment by Alex Miller [ 17/Mar/14 9:48 PM ]

Thanks - please don't use the labels "patch" or "test" - those are covered by the Patch field.

Comment by Devin Walters [ 18/Mar/14 9:17 AM ]

Ah, sorry for the mixup Alex. I assumed you removed "patch" as a label the first time around to flag this ticket as still needing a vetted patch. My mistake.

Comment by Andy Fingerhut [ 21/Mar/14 1:42 PM ]

Patch 0001-Add-transient-predicate.patch dated Mar 17, 2014 applies cleanly to latest Clojure master, but fails a test because the new function transient? has no :added metadata. See most other Clojure functions in clojure.core for examples.

It also generates a new warning while running tests:

WARNING: transient? already refers to: #'clojure.core/transient? in namespace: clojure.test-clojure.data-structures, being replaced by: #'clojure.test-clojure.data-structures/transient?

There is an older (but equivalent) definition of transient? in test file data_structures.clj that should be removed when adding it to clojure.core

Comment by Devin Walters [ 22/Mar/14 11:29 PM ]

@Andy, the reason I did not add :added metadata is because I do not know if/when this patch will be accepted, and as a result, I don't really know if it will sneak into 1.6.X or 1.7. For now, I've put it in as 1.7. If it's in the running to be added sooner than that, let me know and I'll adjust it.

RE: The warning. Good catch. I've submitted a new patch which removes the private version of transient? from data_structures.clj. All tests pass.

Edit to Add: The latest patch as of this comment is now 0002-Add-transient-predicate.patch.





[CLJ-1285] Persistent assoc/conj on a transient-created collision node Created: 28/Oct/13  Updated: 11/Nov/13  Resolved: 11/Nov/13

Status: Closed
Project: Clojure
Component/s: None
Affects Version/s: Release 1.5
Fix Version/s: Release 1.6

Type: Defect Priority: Critical
Reporter: Christophe Grand Assignee: Unassigned
Resolution: Completed Votes: 4
Labels: transient

Attachments: File persistent-assoc-after-collision.diff     File transient-generative-test.diff    
Patch: Code and Test
Approval: Ok

 Description   

Bug reported by Zach Tellman https://groups.google.com/d/msg/clojure-dev/HvppNjEH5Qc/1wZ-6qE7nWgJ

Since transients were introduced the invariant array.length == count*2 doesn't hold for HashCollisionNode.
However persistent .without still relies on it.

Hence persistent dissoc on a collision node created by transients fails.

(let [a (reify Object (hashCode [_] 42))
      b (reify Object (hashCode [_] 42))]
      (= (-> #{a b} transient (disj! a) persistent! (conj a))
       (-> #{a b} transient (disj! a) persistent! (conj a))))

returns false.

Patch: persistent-assoc-after-collision.diff

Generative test patch: transient-generative-test.diff

The generative test reliably reproduces the error. It is simpler than the original test that found the bug but tests a series conj/disj/transient/persistent actions on a set. I've included it separately in case we decide not to apply.

Screened by: Alex Miller



 Comments   
Comment by Michał Marczyk [ 29/Oct/13 9:58 PM ]

I can confirm that the patch works for me. As per our #clojure conversation, I've done the ClojureScript port; see CLJ-648.

Comment by Reid Draper [ 29/Oct/13 11:28 PM ]

I've run Zach's original test, as well as my own simple-check test. Both are passing.

Comment by Alex Miller [ 30/Oct/13 9:33 AM ]

I don't suppose we could get a generative test (prob need to use test.generative which is already included) to test this stuff similar to the original test that found the bug?

Very much hoping to get this into 1.6.

Comment by Andy Fingerhut [ 31/Oct/13 10:52 AM ]

Alex, I suspect clojure-dev would reach a much wider audience for your request than a comment on this ticket, which only has 3 watchers, and I don't think many people besides you and I watch the stream of all ticket state changes as they go by.

Comment by Michał Marczyk [ 01/Nov/13 5:19 AM ]

Just wanted to note that this patch, apart from preventing the hash-based collections from failing Zach's test suite, also makes avl.clj collections pass (now that I've released fixes for the two bugs uncovered by the test suite in avl.clj 0.0.9). This provides some cross-validation, I think.

(The built-in sorted collections pass either way, because they don't support transient ops.)

Also, David Nolen has merged the ClojureScript port of the patch.

Comment by Alex Miller [ 01/Nov/13 7:35 AM ]

I'm going to take a crack at repro with test.generative this morning - wish me luck!

Comment by Alex Miller [ 03/Nov/13 10:40 PM ]

Added a simplified version of a test-generative test and marked screened.

Comment by Alex Miller [ 11/Nov/13 11:17 AM ]

Patch was applied to master for 1.6.





Generated at Tue Jul 29 01:58:23 CDT 2014 using JIRA 4.4#649-r158309.