Clojure JIRA

[CLJ-836] BigInt optimization breaks "obvious" math

Fri, 23 Sep 2011 16:00:06 -0500

BigInt optimization seems seriously broken:

user=> (def a 1N)
#'user/a
user=> (* (+ a 10000000000000000) (+ a 10000000000000000))
ArithmeticException integer overflow
clojure.lang.Numbers.throwIntOverflow (Numbers.java:1374)

A BigInt is optimized back to a long and then overflows which is not
what happened in Beta1 and earlier.

[CLJ-824] BigInt math is slow when values of a BigInt are small enough to actually be treated as Longs

Fri, 26 Aug 2011 14:45:33 -0500

When doing math ops on a BigInt, It should default to treating smaller values of BigInts as Longs when possible to improve performance, and fall through to the actual BigInteger math when necessary.

[CLJ-817] print-deftype breaks when fields have dashes

Fri, 1 Mar 2013 12:46:59 -0600

user=> (deftype Foo [bar-quux])
user.Foo
user=> (def x (Foo. 1))
#'user/x 
user=> x
IllegalArgumentException No matching field found: bar-quux for class user.Foo  clojure.lang.Reflector.getInstanceField (Reflector.java:289)
#user.Foo[user=>

Printing records works fine. See this mailing list message for original report.

[CLJ-816] Transient vectors mutations leak into their source persistent vector

Fri, 1 Mar 2013 12:47:00 -0600

(let [pv (vec (range 34))]
(-> pv transient pop! pop! pop! (conj! 42))
(nth pv 31))
; returns 42 instead of 31

[CLJ-815] [Documentation] - Remove unnecessary "Alpha" labels in docstrings

Fri, 2 Sep 2011 09:41:05 -0500

From smfadi@googlemail.com on the Clojure mailing list

Just a quick question (before I forget): Is deftype/defrecord still
alpha? Or is the "Alpha - subject to change" note in the docstrings
just an oversight?

This needs some housekeeping. Here are a list of functions that are currently marked as "Alpha"

clojure.pprint/print-table
clojure.core/add-watch
clojure.core/remove-watch
clojure.core/juxt
clojure.core/transient
clojure.core/persistent!
clojure.core/conj!
clojure.core/assoc!
clojure.core/dissoc!
clojure.core/pop!
clojure.core/disj!
clojure.core/promise
clojure.core/deliver
clojure.core/defrecord (inside core_deftype)
clojure.core/deftype (inside core_deftype)
clojure.reflect/type-reflect
clojure.reflect/reflect

[CLJ-812] print-dup should not be defined for deftypes

Fri, 29 Jul 2011 10:17:42 -0500

deftypes are supposed to be relatively 'clean'. Defining print-dup on them might conflict with their other intended uses - e.g. if you use a deftype and implement IPersistentSet you can't print due to this conflict.

[CLJ-811] Support hinting arg vectors

Fri, 26 Aug 2011 14:45:55 -0500

Currently, return type hints must be on function or var names, e.g.:

(defn ^String foo [])

This is in contrast to primitive return type declarations, which must be on function arg vectors:

(defn foo ^long [])

The latter is preferred because:

Different arities of the same function can be typed differently
All of the type information for a given function hangs off of a single value (the arg vector)

For these reasons, supporting type hints on arg vectors as well is desirable. This would remove the (confusing and purely historic at this point) syntactic difference between type hints and signature declarations (with the current function/var name hinting support to be potentially deprecated and removed in the future).

As a pleasant side effect, this would disambiguate the semantics of var :tag metadata, thereby resolving CLJ-140.

[CLJ-808] In java.io, the make-parents function can't handle files without parent directory.

Fri, 29 Jul 2011 10:17:41 -0500

The make-parents function throws a null pointer exception when given a simple local directory file.

user> (io/make-parents "test")

Thrown class java.lang.NullPointerException

[CLJ-802] mod throws exception on large args

Fri, 27 May 2011 11:14:26 -0500

From http://groups.google.com/group/clojure-dev/browse_thread/thread/09718dc9c253d1ab

user=> (mod 3216478362187432 432143214)
ArithmeticException integer overflow  clojure.lang.Numbers.throwIntOverflow (Numbers.java:1374)

[CLJ-801] Protocols Should Handle Hash Collision

Tue, 21 Jun 2011 07:41:21 -0500

Protocols internal is using min-hash and throwing a collision, similar to the bug fixed in ~~CLJ-426~~.

https://groups.google.com/d/msg/clojure/0qllPii3fJY/Lile5yQPxuIJ

[CLJ-800] defrecord/deftype enhancements from 1.3.0-alpha7

Fri, 27 May 2011 14:20:38 -0500

The defrecord read/print functionality added in clojure-1.3.0-alpha7 based on ticket CLJ-374 provided a baseline semantics for the described functionality. However, there were some short-comings of the implementation that should be fixed, including:

Symbols not supported in the record/type reader form
Type forms are not performant as a result of being handled by print-dup
- This also leads to a situation where hinting is problematic
The boxClass method uses Reflector
Emission of records pulls out k/v's to create new map to build new record. Record itself should be used instead
Factory fn should not restrict definition of record/type with >20 fields (see here)

The semantics of the 1.3.0-alpha7 behavior should remain intact (save for bug fixes that open more functionality).

[CLJ-798] BigInteger is print-dup'able but not readable

Fri, 29 Jul 2011 10:17:42 -0500

Per Rich, "We should never be printing something that can't be read."

user=> (binding [*print-dup* true] (print-str (java.math.BigInteger. "1")))
"1BIGINT"
user=> (read-string *1)
NumberFormatException Invalid number: 1BIGINT  clojure.lang.LispReader.readNumber (LispReader.java:253)

This can either be resolved by removing the print-dup for BigInteger, printing it using the N notation (thus being read in as a BigInt), or adding read support for the BIGINT notation.

[CLJ-797] Longs print-dup to the Long constructor

Fri, 27 May 2011 14:20:36 -0500

user=> (binding [*print-dup* true] (print-str 1.0))
"1.0"
user=> (binding [*print-dup* true] (print-str 1))
"#=(java.lang.Long. \"1\")"

[CLJ-796] Records with fields named "values"

Fri, 1 Mar 2013 12:46:58 -0600

It looks to me like there is a problem in records when the field name "values" is used:

(defrecord Foo [values])

(= (Foo. 1)
(Foo. 1))

;; -> false

[CLJ-795] bit-clear index should be zero based

Fri, 27 May 2011 11:14:25 -0500

As posted to the mailing list: https://groups.google.com/d/topic/clojure/ky_cVkQyhNw/discussion

alpha7 introduced a regression where the bit-clear index is no longer zero based

user> clojure-version
{:major 1, :minor 3, :incremental 0, :qualifier "alpha4"}
user> (bit-clear 3 1)
1
user> (bit-clear 3 0)
2
user> (bit-clear 3 2)
3
user>
user> clojure-version
{:major 1, :minor 3, :incremental 0, :qualifier "alpha7"}
user> (bit-clear 3 1)
2
user> (bit-clear 3 0)
1
user> (bit-clear 3 2)
0

[CLJ-794] Some alpha-7 printdup functions have left-over debug messages.

Fri, 27 May 2011 11:14:25 -0500

(defmethod print-dup clojure.lang.IPersistentCollection [o, ^Writer w] (print " ipcpd ")
(defmethod print-dup clojure.lang.PersistentHashMap [o w] (print " phmpd ") (print-method o w))

[CLJ-789] Method resolution does not select exact signature matches over tying

Fri, 27 May 2011 11:14:25 -0500

Given 3+ Java methods:

public static boolean someMethod (int a, int b) {
    return true;
}

public static boolean someMethod (int a, long b) {
    return true;
}

public static boolean someMethod (long a, long b) {
    return true;
}

and a Clojure call that matches one of them exactly:

(SomeClass/someMethod (long 1) (long 1))

the compiler yells at us:

More than one matching method found: someMethod

It turns out that Compiler.getMatchingParams is not clearing its "tied" flag when a later signature is an exact match.

[CLJ-784] Update min/max functions to take advantage of primitive math

Fri, 1 Mar 2013 12:47:08 -0600

Currently, the min and max functions are quite slow, since they don't take advantage of new primitive math.

Implement them in the same manner in which other primitive math functions are implemented for efficiency.

[CLJ-782] long cast is not checked for Object decimal types

Fri, 6 May 2011 08:49:10 -0500

E.g.:

user=> (*' Long/MAX_VALUE 100M)
922337203685477580700M
user=> (long *1)
-100
user=> (Double/valueOf Double/MAX_VALUE)
1.7976931348623157E308
user=> (long *1)
9223372036854775807

And the numbers.clj test erroneously considers truncation as correct.

[CLJ-780] race condition in reference cache on Java 5

Fri, 1 Mar 2013 12:47:02 -0600

Map.Entry instances can have null values prior to Java 6

Test failure: http://build.clojure.org/job/test.generative/1/org.clojure$test.generative/console

Discussion: http://concurrency.markmail.org/message/pbruo2uxgur6wkoo?q=map%2Eentry+null&page=3#query:map.entry%20null+page:3+mid:z5arnbbzl2k32jda+state:results

Note that the fix must hold val in a local variable, because cache.remove will bomb if the value is null.

[CLJ-777] Release notes for 1.3

Tue, 23 Aug 2011 18:35:45 -0500

Go through issues and commit logs and assemble Release Notes

http://dev.clojure.org/display/doc/1.3

[CLJ-774] Message-bearing assert

Fri, 29 Apr 2011 10:34:36 -0500

assert should take a String as a second argument. The string is printed when the assert fails.

This is lower hanging fruit than CLJ-415 but much better than the assert we currently have.

In the future, the second argument might be an expression that gets evaluated, but for now, it shouldn't be.

[CLJ-772] bit ops to have primitive semantics by default, no conditionals, direct mapping to JVM primitive ops

Fri, 6 May 2011 08:49:10 -0500

Per Rich's comment on ~~CLJ-767~~.

[CLJ-769] partition-by holds references to head groups longer than necessary

Tue, 21 Jun 2011 14:06:41 -0500

partition-by continues to hold references to items of the input sequence where it should not anymore. That is after (rest s) there are still references held to the head group of the partition-by output sequence.

[CLJ-759] 1.3 alpha 6 seems to hold on to head in a case when 1.2 does not

Tue, 14 Feb 2012 01:06:21 -0600

File names here refer to attached files. Instead of attaching a large input file, I attach a small program that can be run to generate a large input file. To generate it, edit make-input-file.sh to point to a Clojure 1.2 JAR. Also edit run.sh to point to a Clojure 1.2 and 1.3 alpha 6 JAR. Do the following one time to create a ~ 125 Mbyte file long-input.txt:

./make-input-file.sh

With the JVMs I tested, I got these results, where the first argument is 1.2 or 1.3 for the Clojure version, and the second argument is the number of Mbytes to use for the max heap size of the JVM (i.e. the numeric value in the -Xmx<num>m argument in the java command line):

./run.sh 1.2 384 # failed with OutOfMemoryError 5/5 times tried
./run.sh 1.2 416 # succeeded 5/5 times tried

./run.sh 1.3 960 # failed 5/5 times tried
./run.sh 1.3 1048 # succeeded 5/5 or 4/5 times tried, for Mac OS X or Linux, respectively

I believe the issue is that with 1.3 alpha 6, the JVM is holding on to the head of the sequence called 'lines' for the duration of the execution of the function fasta-dna-str-with-desc-beginning, whereas 1.2 is losing the head after getting past the first when-let, and thus using significantly less memory.

[CLJ-756] Workable upload destination for release zips

Fri, 22 Apr 2011 09:34:52 -0500

We have been using Github's file upload under http://github.com/clojure for uploading releases

Pros:

looked easy
download tracking
tool we are already using

Cons:

unreliable
- fails to upload
  - tested on multiple browsers with and without Flash
- even when it works seems to show up during upload
  - presumably would fail the download in these cases
- claims to upload but has garbage file
  - shows up in UI
  - permission error when you try to get it
  - sometimes causes browser to crash (not Github fault, but still...)
- has added two hours of dev time to the current release (so far)

In order to get 1.3.0-alpha6 out the door (and end a multihour ordeal) I am pushing it to the files section at clojure.org.

[CLJ-755] automate distribution zip

Fri, 29 Jul 2011 10:08:40 -0500

In the move to maven, we lost the old ant automation for creating a release zip. In order to release 1.3.0-alpha6, I am using the following ant XML I hacked together:

<target name="dist">
    <property name="clojure.version.label" value="clojure-1.3.0-alpha6"/>
    <property name="distdir" value="dist/${clojure.version.label}"/>
    <mkdir dir="${distdir}"/>
    <copy todir="${distdir}" includeEmptyDirs="false">
      <fileset dir="${basedir}">
        <include name="*.xml"/>
        <include name="**/*.html"/>
        <include name="**/*.txt"/>
        <include name="**/*.markdown"/>
        <include name="**/*.clj"/>
        <include name="**/*.java"/>
      </fileset>
    </copy>
    <copy file="clojure.jar" todir="${distdir}"/>
    <zip basedir="dist" destfile="clojure-${clojure.version.label}.zip"/>
  </target>

What I need is something like this, but properly parameterized and integrated into the maven build, so I can grab it from hudson or central or somewhere after a release build.

[CLJ-752] Remove earmuff == dynamic support

Fri, 1 Mar 2013 12:47:01 -0600

As a compatibility bridge, when dynamic var support was first added, *earmuffed* vars were automagically considered dynamic. Users were promised this bridge would be removed before release. Let's get this change in so people can start dealing with it.

[CLJ-751] cl-format: ~( thows an exception with an empty string

Fri, 1 Mar 2013 12:47:00 -0600

cl-format: ~( thows an exception with an empty string

[CLJ-749] reference a definterface in file that declares it

Fri, 11 Mar 2011 10:21:20 -0600

~~CLJ-737~~ introduced a regression: you can no longer refer to a definterface (e.g. as a type hint) later in the file that declares it. This is caused by the use of Java reflection, where the original code path used ASM reflection.

The patch will keep most of the enhancement of ~~CLJ-737~~ (arrays of primitives), but will not support arrays of classes. (This did not trivially work, but can be made to work as a separate enhancement if anybody strongly needs it.)

[CLJ-748] fast path equal case for diff

Wed, 2 Mar 2011 06:39:56 -0600

For the use case of clojure.data/diff in comparing test results and expectations, it would be nice to do minimal work if the objects are equal. A simple check is low-hanging fruit.

[CLJ-747] stack overflow diffing large objects

Wed, 2 Mar 2011 06:39:57 -0600

e.g.

(clojure.data/diff (range 50000) (range 50000))))

[CLJ-742] Error message for invalid map literals is not helpful

Fri, 25 Feb 2011 15:06:26 -0600

Literal maps with an odd number of forms give an IndexOutOfBounds exception, which is unhelpful and misleading in determining the source of the error. They should instead return a more specific error, something like "map literal must contain an even number of forms".

[CLJ-741] Stop ISeq from inheriting Sequential

Fri, 11 Mar 2011 10:21:10 -0600

Currently, ISeq extends Sequential, which is what collections like vector use to determine equality partition membership. This prevents anything that implements ISeq from being exclusively in the map or set equality partition.

"Care will be required in making sure all appropriate concrete derivees remain Sequential. It's a breaking change in that some derivees not in Clojure may be relying on the derivation from ISeq"

http://groups.google.com/group/clojure-dev/browse_thread/thread/c58ec42d78209ee0

[CLJ-739] version.properties file is not closed

Fri, 25 Feb 2011 15:06:28 -0600

Originally reported by Olek at https://groups.google.com/d/topic/clojure/jhdSyljImuU/discussion

When loading core.clj, the stream opened on the version.properties file is never closed. This can cause problems in some environments such as JEE.

[CLJ-738] <= is incorrect when args include Double/NaN

Fri, 1 Mar 2013 12:47:10 -0600

user> (<= 10 Double/NaN 1)
true

(on both 1.2 and 1.3 alpha 5).

[CLJ-737] definterface/gen-interface do not support array parameter and return types

Fri, 25 Feb 2011 15:25:30 -0600

As describe in this clojure-dev post, gen-interface and definterface do not properly support array type hints.

Ultimately, this is because gen-class and gen-interface use two different code paths to do essentially the same thing. Boiled down, gen-class converts hints using (Type/getType (the-class c)), whereas gen-interface uses asm-type, which uses similar, but different, logic than the-class.

In my patch, I change asm-type to match use the-class. I also add entries to the prim->class map to support primitive arrays.

[CLJ-736] Docstring of defrecord (and =) does not correctly describe equality behavior for records.

Fri, 1 Mar 2013 12:47:03 -0600

http://groups.google.com/group/clojure/browse_frm/thread/eba691b38c45196b#

The docstring of defrecord says it "will define type-and-
value-based equality and hashCode". In reality, record types are included in = but not .equals (or hashCode), and so records of different types can collide as map keys.

Along the same lines, the docstring of = says "same as Java x.equals, except it ... compares numbers and collections in a type-independent manner". To me, this seems to imply the opposite of what actually happens with records.

FWIW I think it would be more clear if the behavior for = and .equals were the same in this respect, but in any case the implemented behavior should be properly documented.

[CLJ-734] starting scope of let bindings seems incorrect from jdi perspective

Fri, 11 Mar 2011 10:21:22 -0600

It appears like the clojure compiler doesn't get the starting scope of let bindings right from the java debug interface perspective.

The compiler sets the starting scope of a local to the let/loop body, rather than to immediately after the local is bound. So when you are stepping through a let statement in a debugger, you can't eval the already defined bindings until you get to the let body.

Because it is a compiler problem, you see the symptoms with any jdi based debugger, (I've confirmed this on both jdb and cdt.)

The fix is pretty straightforward. Just gen.mark() a label after each binding is emitted and use that in the gen.visitLocalVariable() call instead of the loopLabel. I've attached a patch for clojure 1.2 branch. (If there is interest I'll create one for clojure master.)

[CLJ-729] Add any-pred and every-pred combinators

Sun, 20 Mar 2011 08:46:33 -0500

Per the discussion at http://groups.google.com/group/clojure-dev/browse_frm/thread/899349c6a9b526e0

There is a general interest in a set of functions named any/every-pred that take a set of predicates and return a function that applies logical AND/OR to the application of each of its args. The naive implementations would be:

(defn every-pred [& preds] 
  (fn [& args] (every? #(every? % args) preds))) 

((every-pred pos? even?) 1 3 5 7 9)
;=> false

((every-pred pos? odd?) 1 3 5 7 9)
;=> true

(defn any-pred [& preds] 
  (fn [& args] (some #(some % args) preds)))

((any-pred pos? even?) -1 2 3 4 5 6)
;=> true

However, these implementations fall down in the following ways and should be corrected before inclusion to clojure.core:

1) The functions returned by each do not adhere to the same results as (and) and (or) given no arguments.

2) They and their returned functions assume varargs in every call. should be variadically unrolled in much the same way as juxt and comp. A longer term solution would be to implement them in terms of a macro that performs the unrolling automatically, but that is "extra-credit".

3) They should be tested in other_functions.clj

[CLJ-728] Test for error message fails in IBM JDK

Fri, 25 Feb 2011 15:06:28 -0600

A test for reify looks for an error message matching a specific regex, which fails on IBM JDK:

FAIL in (reify-test) (
you can't define a method twice
expected: (fails-with-cause? java.lang.ClassFormatError #"^(Repetitive|Duplicate) method name" (eval (quote (reify java.util.List (size [_] 10) java.util.Collection (size [_] 20)))))
actual: #<CompilerException java.lang.ClassFormatError: (clojure/test_clojure$eval17188$reify__17189) duplicate method at offset=1052, compiling/var/lib/hudson/jobs/ibm-jdk-clojure-build/workspace/src/script/run_tests.clj:257)>

[CLJ-719] clojure.data/diff does not correctly handle arrays as first argument

Fri, 28 Jan 2011 09:24:22 -0600

The EqualityPartition implementation correctly branches, placing arrays into the sequential partition. The Diff implementation needs a similar branch.

[CLJ-716] Hudson release build failing at git clone step

Fri, 1 Mar 2013 12:47:05 -0600

See http://build.clojure.org/job/clojure/237/console for specifics.

[CLJ-715] pprint test failing only on Hudson

Fri, 25 Feb 2011 15:06:28 -0600

The commit https://github.com/clojure/clojure/commit/5a86d525f10f32304f2cb348c9383934a44e6d4b comments out a pprint of agents test that fails only on Hudson. May be a race condition.

You can see the Hudson console log at http://build.clojure.org/job/clojure/236/console (no login required).

[CLJ-710] clojure/set fns don't work with mutable sets

Fri, 14 Jan 2011 07:25:39 -0600

You can't call union, intersection, difference, et al on a mix of Clojure sets and Java sets, because internally the fns assume that sets are clojure values that you can call conj, disj, etc. on.

This percolates up and causes problems for clojure.data/diff as well, since diff aspires to work for any collection, but uses clojure.set to deal with sets.

Possible solutions:

clojure.set fns detect mutable sets and convert to values at the outset
clojure.set doesn't do magic conversions for you. But diff, which is less perf-sensitive, can do so. diff makes immutable values out of mutable sets before handing off to clojure.set
Let the caller beware: docstring in clojure.set tells you to use immutable sets.

[CLJ-708] Multi-methods hold onto the head of their arguments

Fri, 11 Mar 2011 10:21:21 -0600

Multi-methods hold onto the head of their arguments when they are invoked. This means that if you invoke a multi-method with a lazy seq that cannot be held in memory all at once, you blow heap.

I'm not sure how best to write a test case for this particular issue, since the easiest way to test it is to run the JVM with a small heap and purposely evoke an OutOfMemoryError, so the attached patch has only the code changes. (However, I have verified the fix using a small heap.)

This will fix the issue for arities up to 6, for arities >=7 there is a bug in RestFn where it also holds the head of its arguments. If it is desirable to fix that bug as well, then I can submit a patch for it.

[CLJ-702] case gives NPE when used with nil

Fri, 25 Feb 2011 15:06:28 -0600

This code gives a NullPointerException:

(case nil
nil "foo")

If fixing this is impossible for implementation reasons, at least the documentation and error messages should be improved.

[CLJ-697] Compiler doesn't push a binding for unchecked-math

Fri, 1 Mar 2013 12:47:10 -0600

Compile.java in Clojure doesn't push a binding for unchecked-math, so compilation of code with statements like (set! unchecked-math true) fails with:

Can't change/establish root binding of: unchecked-math with set

This problem doesn't happen with the REPL because with-bindings in clojure.main sets up an unchecked-math binding.

[CLJ-695] pprint does not respect print-length

Fri, 1 Mar 2013 12:47:10 -0600

pprint does not respect print-length – instead it prints ellipses for every element one length is passed (and prints forever on infinite collections).

[CLJ-693] VerifyError with symbol metadata, macros, and defrecord

Fri, 1 Mar 2013 12:47:01 -0600

The code below defines a macro wrapper around defrecord. Using it causes a VerifyError:

(defmacro mac1 [name properties] 
  (let [key-info (keyword (first (filter #(meta %) properties)))] 
    (prn key-info) 
    `(defrecord ~name ~properties))) 

(mac1 One [^:key one, two])

Once this happens, the running Clojure process is oddly damaged. Changing the macro to a working version (see below) will not make the sample work with the tagged symbol, almost as if the symbol
'one (below) became corrupt. You either need to start a new REPL, or try to invoke the macro with a different tagged symbol.

The following code does work (note the forced conversion to a string):

(defmacro mac2 [name properties] 
  (let [key-info (keyword (str (first (filter #(meta %) properties))))] 
    (prn key-info) 
    `(defrecord ~name ~properties))) 

(mac2 Two [^:key three, four])

[CLJ-691] missing stacktrace confuses REPL error reporting

Fri, 17 Dec 2010 09:55:38 -0600

when Java is in a bad mood, exceptions start flying without any accompanying stack traces. This causes the default REPL's error reporting to barf, as it currently assumes that the first stack trace element always exists. You can see this by trying the following snippet at the REPL several times in succession:

(def x (byte-array 100000000))

When Java is somewhat angry, you will see the appropriate

OutOfMemoryError Java heap space  clojure.lang.Numbers.byte_array (Numbers.java:1409)

Eventually Java gets angrier and stops providing stack traces, causing an aget error in clojure.main instead.

[CLJ-690] unchecked int math inconsistency with Java

Fri, 8 Apr 2011 09:01:52 -0500

Clojure 1.3.0-alpha4
user=> (-' 0 -9223372036854775808)
-9223372036854775808

[CLJ-689] Audit and update all user permissions on the hudson machine

Fri, 1 Mar 2013 12:47:08 -0600

Currently, the small handful of users are all hudson administrators. With Stuart Sierra's pending changes to the build process, the hudson box will be doing all releases of Clojure. This means that only a few people should be able to run the -alpha, -beta, and stable releases.

[CLJ-686] create JIRA workflow picture

Fri, 22 Apr 2011 09:39:48 -0500

... and post and link from appropriate places

[CLJ-684] agent self-send test heisenfailing

Sun, 19 Dec 2010 15:21:10 -0600

The commented-out test (https://github.com/clojure/clojure/commit/605944b7f667e9fdcc2a380c5dd07304118dca34) is failing intermittently. If you are working on this, look at the changes made in ~~CLJ-672~~, which may be related.

[CLJ-682] cl-format: ~w throws an exception when not wrapped in a pretty-writer

Fri, 1 Mar 2013 12:47:10 -0600

(cl-format nil "hello: _w%" '(a b c))

throws an exception because ~w is (kind of ironically) not marked :pretty in the cl-format definition.

[CLJ-681] Build Clojure with Maven 2

Fri, 1 Mar 2013 12:47:02 -0600

See Common Contrib Build wiki page and the lengthy clojure-dev discussion

For examples of the CI/release process under Hudson, see clojure-testbuild

[CLJ-680] printing promises should not block

Wed, 8 Dec 2010 15:50:13 -0600

Do not tackle the bigger "abstractions around visibility, delay, blocking" etc. issue. This is a minimal fix:

print code already deals with future as a special case
should special-case promise as well
make a definterface IPromiseImpl that you can switch on
- right above promise in core.clj
- has a method that gets you what you need to ask "has-value?"
in print code
- if has value, show it, if not, show :not-delivered
make sure to fix print code in both core_print and pprint

[CLJ-678] into-array should work with all primitive types

Fri, 17 Dec 2010 16:11:24 -0600

e.g.

(into-array Integer/TYPE [0])
(into-array Byte/TYPE [0])

which currently fail with an argument type mismatch.

[CLJ-674] Typo and/or out-of-place example in clojure.string docstring.

Wed, 8 Dec 2010 15:52:29 -0600

Clojure String utilities

It is poor form to (:use clojure.string). Instead, use require
with :as to specify a prefix, e.g.

(ns your.namespace.here
(:require '[clojure.string :as str]))

-----------

The quote in the :require clause is erroneous. On the other hand, clojure.string's ns docstring may not be the place for an explanation of use-vs-require, so the whole paragraph could be removed.

[CLJ-673] RT.load doesn't work when called by code that is on the bootstrap classpath

Fri, 23 Sep 2011 16:00:02 -0500

RT.baseLoader assumes Compiler.getClass.getClassLoader will return a sane value. If Clojure is placed on the bootclasspath, it will return null.

Relying on baseLoader less may be possible; it looks like the ClassLoader/getSystemResource static method may get us the access we need to the bootstrap classloader, which would solve the problem I first noticed with using RT.load with the bootstrap classloader. However, there may be other issues beyond this.

More details at http://groups.google.com/group/clojure/browse_thread/thread/16c694573bd29552

I plan on investigating this further, but I've created this ticket to explain what I've discovered so far for posterity, etc.

Note that placing Clojure on the bootclasspath cuts its startup time in half; it's a measure that Charles Nutter of the JRuby project recommends.

[CLJ-671] Recur mismatch causes infinite loop in compiler

Fri, 27 May 2011 11:14:25 -0500

In cases like

(set! warn-on-reflection true)

(defn gcd [x y]
(loop [x (long x) y (long y)]
(if (== y 0)
x
(recur y (rem x y)))))

I'd expect either both x and y to be auto-boxed or an error from the primitive local mismatch.
Instead the compiler (1.3.0-alpha2) goes into an infinite loop in LetExpr.Parser.parse():

NO_SOURCE_FILE:3 recur arg for primitive local: y is not matching primitive, had: Object, needed: long
Auto-boxing loop arg: y
NO_SOURCE_FILE:3 recur arg for primitive local: x is not matching primitive, had: Object, needed: long
Auto-boxing loop arg: x
NO_SOURCE_FILE:3 recur arg for primitive local: y is not matching primitive, had: Object, needed: long
Auto-boxing loop arg: y
NO_SOURCE_FILE:3 recur arg for primitive local: x is not matching primitive, had: Object, needed: long
Auto-boxing loop arg: x
.
.
.

The previous case has been fixed as described in the comments. The defect can still be seen this way:

(set! warn-on-reflection true)

(defn gcd [x y]
(loop [x (long x) y (long y)]
(if (== y 0)
x
(recur y ^Long(rem x y)))))

[CLJ-665] Add mechanism to temporarily replace root bindings of vars

Fri, 5 Nov 2010 11:05:12 -0500

There are cases, such as mocking out functions during testing, where it is useful to replace the root bindings of vars (dynamic or otherwise) with new values temporarily, and have them reliably set back to their original values when done.

Some discussion here: http://clojure-log.n01se.net/date/2010-10-29.html#09:11

[CLJ-465] with-local-vars broken after changes to make dynamic off by default

Sun, 12 Dec 2010 20:37:22 -0600

with-local-vars should create dynamic vars. Here is what happens now in HEAD

user=> (defn factorial [x]
(with-local-vars [acc 1, cnt x]
(while (> @cnt 0)
(var-set acc (* @acc @cnt))
(var-set cnt (dec @cnt)))
@acc))
#'user/factorial
user=> (factorial 5)
IllegalStateException Can't dynamically bind non-dynamic var: null/null clojure.lang.Var.pushThreadBindings (Var.java:352)

Attached is a patch to mark with-local-vars vars as dynamic, and a test to prove that it works using factorial .

[CLJ-462] clojure.core/slurp docstring out of date

Wed, 8 Dec 2010 15:52:29 -0600

The slurp function used to only work on filename arguments, but now it uses clojure.java.io/reader instead, so it works with a much wider variety of arguments. The docstring does not reflect this change.

Something like this might work: "Opens a reader on f and reads all its contents, returning a string. See clojure.java.io/reader for a complete list of supported arguments."

[CLJ-460] promote interrupt handling to clojure.repl

Fri, 5 Nov 2010 11:05:12 -0500

The functions start-handling-break and add-break-thread! in clojure.contrib.repl-utils are regularly needed by tools such as the Clojure Debugging Toolkit. I propose we promote them to clojure.repl, eliminating what for many projects is their sole dependency on old contrib.

Issues to consider before OKing:

the usage of sun.misc classes appears to be the only game in town. For comparison, JRuby and Groovy already offer access to these classes, Scala appears not to.
the :need-init code in start-handling-break is too cute, but ain't broke, so won't change it.
do we need to add code to clean up the weak references? If so:
- call each time a thread is added?
- switch to ConcurrentHashMap to match code already in Clojure?

[CLJ-458] print-table

Fri, 5 Nov 2010 11:05:12 -0500

Print a textual table from a collection of maps. Useful for viewing regular data at the REPL.

[CLJ-454] docstrings for special ops

Wed, 8 Dec 2010 15:52:29 -0600

"If someone wants to submit a patch with doc support for the special ops, I'll take it, as long as they contain links to the full docs and aren't too long themselves" – rhickey

From here: http://clojure-log.n01se.net/date/2010-10-08.html#10:10

[CLJ-453] Reflection

Fri, 15 Oct 2010 00:30:00 -0500

Clojure API for reflection that

(1) separates API from engine, so that an ASM-based reflector can be used without having to load classes
(2) provides a REPL interface that is a successor to clojure.contrib.repl-utils/show
(3) returns data, not encapsulated objects

[CLJ-448] structural diff

Fri, 5 Nov 2010 15:02:38 -0500

I constantly find myself writing ad hoc code to compare bits of structure, either at the REPL or in tests. (Or even worse, trying to eyeball two long pprint outputs).

What I want instead is a diff function that compares two structures a and b, returning the parts only in a, parts only in b, and parts in both. Good solution for my needs should:

work with all kinds of collections (both Clojure and Java)
make simple choices about how to define a "difference"
need not worry too much about perf

[CLJ-447] biginteger fn misses a case for clojure.lang.BigInt

Fri, 5 Nov 2010 11:05:12 -0500

(biginteger 13178456923875639284562345789N)
==> #<IllegalArgumentException java.lang.IllegalArgumentException: Value out of range for long: 13178456923875639284562345789>

[CLJ-444] Infinite recursion in Keyword.intern leads to stack overflow

Tue, 28 Sep 2010 21:30:00 -0500

Keyword.intern keeps a table of SoftReferences to keywords, always preferring the existing one to making a new one. However, if the SoftReference has been pulled, the code tries to recurse as its way of "trying again."

Unfortunately, the way the logic is written, the code will work exactly the same way the second time around (and the third... it's turtles all the way down the stack) and eventually have a stack overflow.

I found this in a long running web service that was using XML RPC and then "keywordizing" the keys of the result maps. Within an hour or so, it would blow up on this. From my read of the code, there is no work around once you're in this state since the expired reference is stuck in the keyword table.

Patch coming momentarily.

[CLJ-443] Add zero-arity body to comp function

Mon, 11 Oct 2010 14:25:00 -0500

The comp function should return the identity function should it be called with zero args. Currently there is no zero-arity form. I talked a little more about this at http://blog.fogus.me/2010/08/18/monkeying-with-clojures-comp-function/. Patch attached.

[CLJ-442] Fix embedded doc links on clojure.org docs

Mon, 15 Nov 2010 12:49:00 -0600

These broke when we moved to clojure org on github. Try WebDAV for bulk edit without having to troll through the pages manually

[CLJ-441] Add unchecked coercions

Wed, 8 Dec 2010 15:52:28 -0600

Checking in coercions has been added, making them safe, but removing the bit-twiddling power. We need some unchecked (truncating) coercion ops

[CLJ-439] Automatic type translation from Integer to Long

Tue, 28 Sep 2010 13:39:00 -0500

In some cases types can get lost:

(def i1 (Integer. "100"))
(def i2 (Integer/valueOf 100))

[(class i1) (class i2)]
==> [java.lang.Integer java.lang.Long]

;; expected ==> [java.lang.Integer java.lang.Integer]

;; But: (class (Integer/valueOf 100)) ==> java.lang.Integer

Works correspondingly for Floats, which become Doubles.

[CLJ-437] Bugs in clojure.set/subset? and superset? for sets with false/nil elements

Tue, 12 Oct 2010 19:58:00 -0500

clojure.set/subset? and /superset? fail on sets with false or nil elements, since they use the input sets directly as a membership predicates:

user> (clojure.set/subset? #{nil} #{nil 2 3})
false

http://groups.google.com/group/clojure-dev/browse_frm/thread/c9c2bebd64803959/e61bbc0d148ecd7b#e61bbc0d148ecd7b

[CLJ-435] stackoverflow exception in printing meta with :type

Fri, 29 Apr 2011 10:34:37 -0500

Hi, after expose the problem in clojure irc and the help of @chouser, we thought my problem must be a bug.
Here the console session

user> (with-meta {:value 2} {:type Object})

No message. [Thrown class java.lang.StackOverflowError]

0: clojure.lang.PersistentHashMap$BitmapIndexedNode.index(PersistentHashMap.java:485)
1: clojure.lang.PersistentHashMap$BitmapIndexedNode.find(PersistentHashMap.java:571)
2: clojure.lang.PersistentHashMap$ArrayNode.find(PersistentHashMap.java:355)
3: clojure.lang.PersistentHashMap.entryAt(PersistentHashMap.java:129)
4: clojure.lang.Var.getThreadBinding(Var.java:334)
5: clojure.lang.Var.deref(Var.java:137)
6: clojure.lang.Var.get(Var.java:133)
7: clojure.core$pr_on.invoke(core.clj:2810)
8: clojure.lang.Var.invoke(Var.java:369)
9: clojure.lang.RT.print(RT.java:1270)
10: clojure.lang.RT.printString(RT.java:1249)
11: clojure.lang.APersistentMap.toString(APersistentMap.java:20)

Chouser pointed that is a print error, creation and use of map with meta works fine

Thank for your help!

[CLJ-433] munge should not munge $ (which isJavaIdentifierPart), should munge ' (which is not)

Fri, 1 Mar 2013 12:46:59 -0600

munge is overeager in converting $ to DOLLARSIGN, since $ is a valid
character in Java identifiers:

user> (filter #(Character/isJavaIdentifierPart %) (keys
clojure.lang.Compiler/CHAR_MAP))
(\$)

On a related point, ' (single quote) is not admissible in Java
identifiers, but it is not munged on master:

user=> (Character/isJavaIdentifierPart \')
false
user=> (munge "'")
"'"

This leads to e.g.

user=> *'
#<core$_STAR_' clojure.core$_STAR_'@5adf48c4>

(note the ' after STAR).

Originally reported on the Dev list.

See also this thread on the (regular) Clojure ggroup.

The attached patch applies cleanly against current master.

[CLJ-432] deftype does not work if containing ns contains dashes

Sun, 13 Mar 2011 23:54:53 -0500

There's a problem with the compilation (either live or AOT) of e.g. types, if the namespace containing the type definition cointains dashes.

It's quite easy to reproduce: create src/net/yournick/lr_plus.clj , and in this file, have just "(ns net.yournick.lr-plus) (deftype Foo)"
AOT compile this, and shebang! in your classes/ directory you see the following folder structure: classes/net/yournick/lr-plus/

lr-plus/ contains the Foo.class file

folder lr-plus/ should really be lr_plus/

the problem is that while apparently with Oracle JVMs everything works fine while you don't try to AOT compile it, it does not work (even if not AOT'ed) with IBM JRE 6:

Caused by: java.lang.ClassFormatError: JVMCFRE068 class name is invalid; class=compile__stub/net/cgrand/parsley/lr-plus/TableState, offset=0 (lr_plus.clj:8)

[CLJ-430] clojure.java.io URL Coercion throws java.lang.ClassCastException

Mon, 11 Oct 2010 14:25:00 -0500

I noticed a bug in the clojure.java.io namespace at Line 57 in extend-protocol Coercions URL

URL
   (as-url [u] u)
   (as-file [u]
     (if (= "file" (.getProtocol u))
       (as-file (.getPath u))
--      (throw (IllegalArgumentException. "Not a file: " u))))
++      (throw (IllegalArgumentException. (str "Not a file: " u)))))
</code></pre>
Right now it throws the error:
<pre><code>
Caused by: java.lang.ClassCastException: java.net.URL cannot be cast to java.lang.Throwable
        at clojure.java.io $fn__7354.invoke(io.clj:56)
        at clojure.java.io $fn__7328$G__7323__7333.invoke(io.clj:34)

Clojure version: 1.2

[CLJ-429] revamp ant build

Fri, 1 Oct 2010 08:53:00 -0500

reduce number of manual steps, improve clarity, add options

[CLJ-427] Github URL on clojure.org submission guidelines is out of date

Tue, 28 Sep 2010 16:00:00 -0500

Need to update the git clone URLs on this page: http://clojure.org/patches as it still points to git://github.com/richhickey/clojure-contrib.git.

[CLJ-426] case should handle hash collision

Fri, 29 Apr 2011 10:34:39 -0500

should generate a branch under the hash to deal with the conflicting values

[CLJ-425] stop early-loading ancillary libraries

Wed, 22 Sep 2010 14:25:00 -0500

In RT.java, doInit loads a bunch of libraries on startup:

load("clojure/zip", false);
 load("clojure/xml", false);
 load("clojure/set", false);

This is vestigial from before use/require, and Clojure would load faster if it went away. Would be a small breaking change if anyone uses these libraries with full namespace qualification, and without a prior load/use.

[CLJ-421] clojure tests can fail because of a race in the pprint tests

Tue, 28 Sep 2010 13:00:00 -0500

When I was doing the pprint tests for printing agents, I typod the code that guarded against races.

Bug was observed by Rasmus Svensson in http://groups.google.com/group/clojure-dev/browse_thread/thread/26b14bcada913247#

Example:

[java] FAIL in (pprint-datastructures-tests) (test_pretty.clj:226)
[java] expected: ((var
clojure.test-clojure.pprint.test-helper/back-match) (tst-pprint 20
(failed-agent)) #"#<Agent@[0-9a-f]+ FAILED: \n \"foo\">")
[java] actual: nil

Patch coming momentarily.

[CLJ-410] clojure.xml parse/emit do not round-trip

Mon, 15 Nov 2010 12:49:00 -0600

I'm not expecting (emit (parse x)) to produce the same characters as originally in x. (That would be canonical form.)
However, it seems reasonable to expect that (= (parse x) (parse (emit (parse x)))). That's not currently the case because of #277 and #408.
It's not currently possible to demonstrate this bug via a unit test because of #409.
This fails:

(deftest parse-and-emit-should-round-trip-data
  (are [s0]
       (let [x1 (str->xml s0)
             s2 (xml->str x1)
             x3 (str->xml s2)]
         (is (= x1 x3)))
       "<e/>"
       "<e></e>"
       "<e>content</e>"
       "<e a='attribute'/>"
       "<e a='attribute'>content</e>"
       "<e><e><e/></e></e>"
       "<e> </e>"
       "<e>&lt;&amp;&gt;</e>"))
</code></pre>
Where:
<pre><code>(defn xml->str [x]
  (with-out-str
    (xml/emit-element x)))

(defn str->xml [s]
  (xml/parse
   (ByteArrayInputStream. (.getBytes s "UTF-8"))))

Or, rather it crashes prematurely in str->xml because of #409.

[CLJ-397] better error message when calling macros with arity

Fri, 5 Nov 2010 11:05:12 -0500

The two magic arguments to macros lead to weird error messages:

(defmacro foo [a b] a)
(foo)
java.lang.IllegalArgumentException: Wrong number of args (2) passed to: user$foo (NO_SOURCE_FILE:)0

[CLJ-390] sends from agent error-handlers should be allowed

Wed, 8 Dec 2010 15:52:29 -0600

Currently if an agent's error-handler sends to any agent, the sent action is
silently thrown away. The send should instead be allowed.

Originally reported here:
http://groups.google.com/group/clojure/browse_frm/thread/c1d05bdbb50d7d28

[CLJ-380] bit-and missing long parameters overload

Fri, 25 Feb 2011 15:06:28 -0600

[Updated: The new primitive support relies on a primitive bit-and operator. As one doesn't exist, clojure functions like even? are now using reflective calls and are about 150x slower than in Clojure 1.2)

JVM has a land instruction and java has & operator that works with longs.

public static long and(long, long);
Code:
0: lload_0
1: lload_2
2: land
3: lreturn

The attached patch allows Clojure to take advantage of it.

[CLJ-378] Set thread names on agent thread pools

Tue, 12 Oct 2010 19:58:00 -0500

It��s a best practice to name the threads in an executor thread pool with a custom ThreadFactory so that the purpose of these threads is clear in thread dumps and other runtime operational tools. By default these threads are currently called something like "pool-%d-thread-%d", and this is what you��ll see for the agent send thread pools.

I created a patch to do this with thread names like:

clojure-agent-send-pool-%d �� should be fixed # of threads
clojure-agent-send-off-pool-%d �� will be added and removed over time

The patch is attached and I have a signed CA.

[CLJ-374] print/read syntax for defrecords

Fri, 27 May 2011 11:14:25 -0500

print/read syntax for defrecords

[CLJ-368] Redefining a function that uses a redifined macro, picks up old macro defintion

Mon, 11 Oct 2010 15:25:00 -0500

Redefining a function to use an updated macro definition fails to work correctly, with the function still using the previous macro definiton. This seems to be a regression.

Test case: http://gist.github.com/420007
Bisection : http://gist.github.com/420075

[CLJ-364] Make PersistentQueue Counted

Fri, 15 Oct 2010 13:30:00 -0500

Currently, calling count() on PersistentQueue can take a long time for large queues. PersistentQueue should implement the Counted interface and provide constant time counting by maintaining a count of the number of elements internally.

[CLJ-336] top-level nav for clojure.org

Mon, 15 Nov 2010 12:49:00 -0600

top-level nav for clojure.org

[CLJ-320] alias function gives confusing message if using unknown namespace

Wed, 8 Dec 2010 15:52:29 -0600

This is really a request for improved error handling. If you use (alias) with a namespace that is not known it throws an NPE with imho a confusing error message.

Example:

user=> (alias 'R 'S) 
java.lang.NullPointerException: Expecting Symbol + Namespace (NO_SOURCE_FILE:0)
user=> (.printStackTrace *e)
nil
user=> java.lang.NullPointerException: Expecting Symbol + Namespace (NO_SOURCE_FILE:0)
        at clojure.lang.Compiler.eval(Compiler.java:5348)
        at clojure.lang.Compiler.eval(Compiler.java:5300)
        at clojure.core$eval__4281.invoke(core.clj:2135)
        at clojure.main$repl__6358$read_eval_print__6366.invoke(main.clj:183)
        at clojure.main$repl__6358.doInvoke(main.clj:200)
        at clojure.lang.RestFn.invoke(RestFn.java:422)
        at clojure.main$repl_opt__6392.invoke(main.clj:254)
        at clojure.main$main__6418.doInvoke(main.clj:347)
        at clojure.lang.RestFn.invoke(RestFn.java:398)
        at clojure.lang.Var.invoke(Var.java:361)
        at clojure.lang.AFn.applyToHelper(AFn.java:161)
        at clojure.lang.Var.applyTo(Var.java:482)
        at clojure.main.main(main.java:37)
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
        at java.lang.reflect.Method.invoke(Method.java:597)
        at jline.ConsoleRunner.main(ConsoleRunner.java:69)
Caused by: java.lang.NullPointerException: Expecting Symbol + Namespace
        at clojure.lang.Namespace.addAlias(Namespace.java:184)
        at clojure.core$alias__4566.invoke(core.clj:2892)
        at user$eval__11.invoke(NO_SOURCE_FILE:7)
        at clojure.lang.Compiler.eval(Compiler.java:5332)
        ... 17 more
</code></pre>

And the clj:

<pre><code>(defn alias
  "Add an alias in the current namespace to another
  namespace. Arguments are two symbols: the alias to be used, and
  the symbolic name of the target namespace. Use :as in the ns macro in preference
  to calling this directly."
  [alias namespace-sym]
  (.addAlias *ns* alias (find-ns namespace-sym)))

Clearly, the clojure is doing find-ns and the namespace is not known so it passes a null down. Seems like the code in either the clj or java level could say "Hey dummy, you used an unknown namespace" instead of "Expecting Symbol + Namespace". Maybe this would be useful as an additional sentence in the alias doc string too.

[CLJ-315] Add support for running -main namespace from clojure.main without AOT

Fri, 15 Oct 2010 04:30:00 -0500

This patch allows clojure.main to accept an argument pointing to a namespace to look for a -main function in. This allows users to write -main functions that will work the same whether the code is AOT-compiled for use in an executable jar or just run from source.

Discussed on the mailing list: http://groups.google.com/group/clojure/browse_thread/thread/bcf3c7c0e002601e

[CLJ-305] out has a misleading docstring

Fri, 15 Oct 2010 13:30:00 -0500

The docstring for out reads:

Defaults to System/out

Like in, it should read something like:

Defaults to System/out, wrapped in an OutputStreamWriter

[CLJ-300] newline should output platform-specific newline sequence

Fri, 17 Dec 2010 16:11:24 -0600

(newline), and (println) currently output \n as the newline sequence, even on Windows, where \r\n is the newline sequence.
These functions should output the platform specific newline sequence.

[CLJ-292] LazySeq.sval() nests RuntimeExceptions

Wed, 8 Dec 2010 15:50:13 -0600

When evaluating a LazySeq, if an exception occurs it gets wrapped in a RuntimeException and is thrown. Unfortunately, the logic does not differentiate between a RuntimeException and a checked exception. This has what I believe are two unintended side effects.

If a function throws a RuntimeException, such as an IllegalArgumentException, it gets wrapped in a RuntimeException unnecessarily. So, if I am expecting an IllegalArgumentException (as in a test), I will unexpectedly get a RuntimeException instead.
When evaluating nested lazy sequences, if an error occurs, the root exception is wrapped over and over again as the exception winds its way up the stack. In a recent project, this meant having extremely long stack traces that include a half dozen nested RuntimeExceptions. These just get in the way of getting to the root cause.

Both of these scenarios are illustrated in the attached file.

My proposed solution is fairly simple: Instead of indiscriminately wrapping every exceptions a RuntimeException, simply rethrow any RuntimeException. As a result, any RuntimeException will simply wind its way back up the stack with no wrapping, and checked exceptions will only be wrapped once. This should make debugging Clojure programs slightly easier as the resulting stack trace will not include as much irrelevant information.

[CLJ-286] out being used as java.io.PrintWriter

Thu, 14 Oct 2010 14:27:00 -0500

There seem to be several places in clojure/contrib where out is being used as a java.io.PrintWriter, despite being documented as of type java.io.Writer
Examples are:

The reflection warnings from the compiler
Exception printouts from contrib.sql

This causes very confusing errors, when out isn't actually a PrintWriter, for example in a swank repl where it is a StringWriter.

I've created a patch for the reflection warnings, but can't submit it here because of political issues.

[CLJ-285] Make it easier to discover the source of compiler exceptions when two files have the same name in different directories

Mon, 11 Oct 2010 07:25:00 -0500

Compiler exceptions only print the source name of exceptions, not the path. If you have a source tree with the files a/b/foo.clj and c/d/foo.clj, you'll get error messages like Error: no such namespace (foo.clj: 100). The attached patch modifies Compiler.java to print SOURCE_PATH rather than SOURCE in several cases.

[CLJ-282] Replace references to duck-streams in clojure.test comments to io

Mon, 11 Oct 2010 08:37:00 -0500

duck-streams was renamed to io back in February, but the references in clojure.test have not been updated. Attached is a patch with this simple fix.

[CLJ-280] def should support an optional doc-string

Fri, 28 Jan 2011 15:02:08 -0600

To be more consistent across all "operators that introduce names", def should support an optional docstring after the name.

[CLJ-278] replicate could be marked deprecated

Mon, 11 Oct 2010 09:25:00 -0500

replicate has become unnecessary, as repeat is exactly similar when called with two arguments.
Therefore, it might be marked deprecated.

[CLJ-276] Add non-interning keyword lookup function

Tue, 12 Oct 2010 16:58:00 -0500

I need a non-interning keyword lookup function. This is the much delayed patch discussed way back, see this discussion:
http://groups.google.com/group/clojure-dev/browse_thread/thread/670ab284e86e17fa/22f3ba493a93db8c?lnk=gst&q=keyword+lookup#22f3ba493a93db8c

Attached is a proposed patch to implement this. It's different from the patch in the above thread in three main ways: it's against a current branch, it includes a clojure wrapper (find-keyword), and it uses the name find in the keyword class instead of lookup.

It's currently interning symbols, which I'm not sure about. It might be better to have the keys of the keyword table be strings instead of symbols, but that's a more intrusive change. Also, the find-keyword function is in core, which I'm not sure is the right place for it either.

[CLJ-263] Add an optional environment argument to #'resolve and ns-resolve

Fri, 15 Oct 2010 05:30:00 -0500

In ns-resolve and resolve add an arity overload with an environment argument before sym.

See http://groups.google.com/group/clojure-dev/msg/445e31ccd0cb6988
and http://clojure-log.n01se.net/date/2010-02-10.html#12:10

[CLJ-249] Add "starting at " to EOF while reading exceptions

Sat, 2 Oct 2010 00:48:00 -0500

Extend some EOF exceptions during read so they look like this:

java.lang.Exception: EOF while reading, starting at line 7 (bad.clj:15)

So while bad.clj:15 indicates the file name and the fact that
line 15 is the last one of the file, "starting at line 7" points
out the line where the unclosed grouping construct started.

Discussed: http://groups.google.com/group/clojure/browse_thread/thread/872c5b2ee7af19fe

[CLJ-244] walk support for sorted-by collections

Fri, 1 Mar 2013 12:47:08 -0600

http://groups.google.com/group/clojure-dev/browse_thread/thread/51982e4091e3614d

[CLJ-236] Add error checking for defmulti options.

Fri, 29 Apr 2011 10:34:37 -0500

It would be nice to have some error checking for option map in macros like defmulti. I've written a check-options function that can be used for that purpose and used it in defmulti definition in the attached patch.

[CLJ-235] clojure.test fixtures don't work with test-ns-hook

Fri, 5 Nov 2010 11:05:12 -0500

I just noticed something about clojure.test that isn't documented, and might not be by design.

The docs say:

"Fixtures allow you to run code before and after tests, to set up
the context in which tests should be run."

and also

"By default, these functions will search for all tests defined in
a namespace and run them in an undefined order. However, if you
are composing tests, as in the "arithmetic" example above, you
probably do not want the "addition" and "subtraction" tests run
separately. In that case, you must define a special function
named "test-ns-hook" that runs your tests in the correct order:

(defn test-ns-hook []
     (arithmetic))

The actual behavior is that fixtures are run if you don't specify your own test-ns-hook</code>. You can figure this out if you read the source, or from the docstring of <code>test-ns if you know how fixtures are implemented, but it's certainly not clear.

Might I suggest that one of the following be done?

The docs are extended or clarified to warn that fixtures will not be run if you define your own test-ns-hook, or
test-ns</code> be extended (and <code>test-all-vars</code> refactored) to call fixtures prior to calling the <code>test-ns-hook, or
An exported function be introduced (and test-all-vars refactored to use it) to thread execution via the ns fixtures, so that hooks can use it?

I'm leaning towards the second or third option; the third allows the test-ns-hook</code> to optionally invoke the fixtures, which adds flexibility, but of course if you're defining a <code>test-ns-hook you're at liberty to simply not define the fixtures at all (which lends weight to the second solution).

Note that it might only make sense to run "once" fixtures around a custom test-ns-hook, unless tests themselves are put in charge of running "each" fixtures.

[CLJ-234] Adding provenance of deprecation warnings to the LispReader

Fri, 1 Mar 2013 12:47:03 -0600

The LispReader has some deprecation warnings, but they do not show their provenance. The attached patch provide improved warning messages by modifying the LineNumberingPushbackReader to include the source path of the file being read, so that it can be retrieved by the LispReader.

[CLJ-199] chunked seqs support 'for' byte code can't be converted to dex for dalvik

Mon, 15 Nov 2010 12:49:00 -0600

Commit 306ef6 generates bytecode which can not be covered to dex for dalvik. See following git-bisect output for details.

306ef6d9e47253103ca0e5ae0f4b620d5fa2aeff is first bad commit
commit 306ef6d9e47253103ca0e5ae0f4b620d5fa2aeff (14316ae2110a779ffc8ac9c3da3f1c41852c4289)
Author: Chouser <chouser@n01se.net>
Date:   Sun Aug 23 21:36:41 2009 -0400

    Add support for chunked seqs to 'for'.  Refs #1

:040000 040000 d48a1b12db408d4bc5843367ea5529678522f57d 7d8717380716feab9ebc84883e60dceea8fb1e08 M src
:040000 040000 0d80fce85701005c13ce13c105524a4fff188396 5daece53b0dd2727dc7ffa9c0b773e7962de900d M test
bisect run success

processing clojure/core$generate_proxy__5911$iter__5947__5953$fn__5954$iter__5949__5956$fn__5957.class...

UNEXPECTED TOP-LEVEL EXCEPTION:
com.android.dx.cf.code.SimException: local variable type mismatch: attempt to set or access a value of type java.lang.Object using a local variable of type int. This is symptomatic of .class transformation tools that ignore local variable information.
 at com.android.dx.cf.code.BaseMachine.throwLocalMismatch(BaseMachine.java:537)
 at com.android.dx.cf.code.Simulator$SimVisitor.visitLocal(Simulator.java:504)
 at com.android.dx.cf.code.BytecodeArray.parseInstruction(BytecodeArray.java:472)
 at com.android.dx.cf.code.Simulator.simulate(Simulator.java:96)
 at com.android.dx.cf.code.Ropper.processBlock(Ropper.java:681)
 at com.android.dx.cf.code.Ropper.doit(Ropper.java:636)
 at com.android.dx.cf.code.Ropper.convert(Ropper.java:253)
 at com.android.dx.dex.cf.CfTranslator.processMethods(CfTranslator.java:252)
 at com.android.dx.dex.cf.CfTranslator.translate0(CfTranslator.java:131)
 at com.android.dx.dex.cf.CfTranslator.translate(CfTranslator.java:85)
 at com.android.dx.command.dexer.Main.processClass(Main.java:297)
 at com.android.dx.command.dexer.Main.processFileBytes(Main.java:276)
 at com.android.dx.command.dexer.Main.access$100(Main.java:56)
 at com.android.dx.command.dexer.Main$1.processFileBytes(Main.java:228)
 at com.android.dx.cf.direct.ClassPathOpener.processArchive(ClassPathOpener.java:245)
 at com.android.dx.cf.direct.ClassPathOpener.processOne(ClassPathOpener.java:130)
 at com.android.dx.cf.direct.ClassPathOpener.process(ClassPathOpener.java:108)
 at com.android.dx.command.dexer.Main.processOne(Main.java:245)
 at com.android.dx.command.dexer.Main.processAllFiles(Main.java:183)
 at com.android.dx.command.dexer.Main.run(Main.java:139)
 at com.android.dx.command.dexer.Main.main(Main.java:120)
 at com.android.dx.command.Main.main(Main.java:87)
...at bytecode offset 000000d7
locals[0000]: Lclojure/core$generate_proxy__5911$iter__5947__5953$fn__5954$iter__5949__5956$fn__5957;
locals[0001]: known-null
locals[0002]: known-null
locals[0003]: known-null
locals[0004]: known-null
locals[0005]: I
locals[0006]: Ljava/lang/Object;
stack[0003]: Lclojure/lang/IFn;
stack[0002]: Ljava/lang/Object;
stack[0001]: Ljava/lang/Object;
stack[top0]: known-null
...while working on block 00cd
...while working on method invoke:()Ljava/lang/Object;
...while processing invoke ()Ljava/lang/Object;
...while processing clojure/core$generate_proxy__5911$iter__5947__5953$fn__5954$iter__5949__5956$fn__5957.class

[CLJ-191] enhance stacktrace: causes like java, syntax like clojure

Fri, 1 Mar 2013 12:47:08 -0600

Various enhancements I'd like to make:
Copy the feature of java stacktrace to print "... 25 more" instead of duplicating stack lines.
Print java class.method in clojure format: class/method.

Discussion: http://groups.google.com/group/clojure-dev/browse_thread/thread/d03dd16acbe3aa14

[CLJ-184] n-ary bit functions, also inlining of n-ary bit and math operations

Fri, 1 Mar 2013 12:47:03 -0600

Add an &more version to the bit operations bit-and bit-or bit-xor bit-and-not (via reduce),
Inline the n-ary version of the math ops, + - * /, inline the n-ary versions of bit-and bit-or bit-xor bit-and-not

[CLJ-182] Make Symbol.create methods intern their args

Thu, 14 Oct 2010 14:27:00 -0500

Right now the create methods presume interned strings, but not all users are disciplined about that. Also, Terracotta can correctly identity-share explicitly interned strings but not implicit ones (like string literals). We care because symbol equality is based on string identity.

[CLJ-149] pop does not preserve meta-data

Fri, 1 Mar 2013 12:47:08 -0600

As discussed with chouser on irc.

Example:

=> (meta (conj (with-meta '(1 2 3) {:my :meta}) 4))
{:my :meta}
=> (meta (pop (with-meta '(1 2 3) {:my :meta}))) 
nil

Both calls should yield the same result.

This occurs in 1.0.0 as well as trunk.

This might be part of a larger issue with when meta-data should be preserved (e.g., butlast).

[CLJ-139] defmulti should not add {:tag 'clojure.lang.MultiFn} to multimethod metatdata

Mon, 11 Oct 2010 15:25:00 -0500

defmulti's current behavior of adding a {:tag 'clojure.lang.MultiFn} to the new multimethod's metadata interferes with providing a type hint for the return value of the multimethod. This is easily fixed by deleting this behavior, as in the attached patch.

[CLJ-31] GC Issue 27: Disallow recur across try

Sat, 18 Dec 2010 08:41:40 -0600

Reported by richhickey, Jan 01, 2009
Right now will allow:

(loop [x 42]
  (when (pos? x)
    (binding [*out* *out*]
      (recur (dec x)))))

but the recur doesn't (can't) really happen in the scope of the binding

[CLJ-8] GC Issue 3: Re-enable detection of circular loads

Fri, 28 Jan 2011 09:24:22 -0600

Reported by richhickey, Dec 17, 2008
Was made a no-op for AOT/gen-class - need to distinguish cases

Comment 1 by richhickey, Dec 17, 2008
(No comment was entered for this change.)
Owner: ---
Comment 2 by richhickey, Dec 18, 2008
(No comment was entered for this change.)
Labels: Priority-High
Comment 3 by scgilardi, May 12, 2009
Rich, do you have an example of the problem with AOT/gen-class that motivated
disabling detection of circular loads? With that in hand, I'd like to take a look at
fixing this.
Comment 4 by richhickey, May 13, 2009
You should just try it out. I'm not sure if it is just something I encountered on my
way to the final mechanism (and now isn't needed). However, you should hold off right
at the moment, as I'm working on loading/classloader stuff.

Clojure JIRA

[CLJ-836] BigInt optimization breaks "obvious" math

[CLJ-824] BigInt math is slow when values of a BigInt are small enough to actually be treated as Longs

[CLJ-817] print-deftype breaks when fields have dashes

[CLJ-816] Transient vectors mutations leak into their source persistent vector

[CLJ-815] [Documentation] - Remove unnecessary "Alpha" labels in docstrings

[CLJ-812] print-dup should not be defined for deftypes

[CLJ-811] Support hinting arg vectors

[CLJ-808] In java.io, the make-parents function can't handle files without parent directory.

[CLJ-802] mod throws exception on large args

[CLJ-801] Protocols Should Handle Hash Collision

[CLJ-800] defrecord/deftype enhancements from 1.3.0-alpha7

[CLJ-798] BigInteger is print-dup'able but not readable

[CLJ-797] Longs print-dup to the Long constructor

[CLJ-796] Records with fields named "values"

[CLJ-795] bit-clear index should be zero based

[CLJ-794] Some alpha-7 printdup functions have left-over debug messages.

[CLJ-789] Method resolution does not select exact signature matches over tying

[CLJ-784] Update min/max functions to take advantage of primitive math

[CLJ-782] long cast is not checked for Object decimal types

[CLJ-780] race condition in reference cache on Java 5

[CLJ-777] Release notes for 1.3

[CLJ-774] Message-bearing assert

[CLJ-772] bit ops to have primitive semantics by default, no conditionals, direct mapping to JVM primitive ops

[CLJ-769] partition-by holds references to head groups longer than necessary

[CLJ-759] 1.3 alpha 6 seems to hold on to head in a case when 1.2 does not

[CLJ-756] Workable upload destination for release zips

[CLJ-755] automate distribution zip

[CLJ-752] Remove *earmuff* == dynamic support

[CLJ-751] cl-format: ~( thows an exception with an empty string

[CLJ-749] reference a definterface in file that declares it

[CLJ-748] fast path equal case for diff

[CLJ-747] stack overflow diffing large objects

[CLJ-742] Error message for invalid map literals is not helpful

[CLJ-741] Stop ISeq from inheriting Sequential

[CLJ-739] version.properties file is not closed

[CLJ-738] <= is incorrect when args include Double/NaN

[CLJ-737] definterface/gen-interface do not support array parameter and return types

[CLJ-736] Docstring of defrecord (and =) does not correctly describe equality behavior for records.

[CLJ-734] starting scope of let bindings seems incorrect from jdi perspective

[CLJ-729] Add any-pred and every-pred combinators

[CLJ-728] Test for error message fails in IBM JDK

[CLJ-719] clojure.data/diff does not correctly handle arrays as first argument

[CLJ-716] Hudson release build failing at git clone step

[CLJ-715] pprint test failing only on Hudson

[CLJ-710] clojure/set fns don't work with mutable sets

[CLJ-708] Multi-methods hold onto the head of their arguments

[CLJ-702] case gives NPE when used with nil

[CLJ-697] Compiler doesn't push a binding for *unchecked-math*

[CLJ-695] pprint does not respect *print-length*

[CLJ-693] VerifyError with symbol metadata, macros, and defrecord

[CLJ-691] missing stacktrace confuses REPL error reporting

[CLJ-690] unchecked int math inconsistency with Java

[CLJ-689] Audit and update all user permissions on the hudson machine

[CLJ-686] create JIRA workflow picture

[CLJ-684] agent self-send test heisenfailing

[CLJ-682] cl-format: ~w throws an exception when not wrapped in a pretty-writer

[CLJ-681] Build Clojure with Maven 2

[CLJ-680] printing promises should not block

[CLJ-678] into-array should work with all primitive types

[CLJ-674] Typo and/or out-of-place example in clojure.string docstring.

[CLJ-673] RT.load doesn't work when called by code that is on the bootstrap classpath

[CLJ-671] Recur mismatch causes infinite loop in compiler

[CLJ-665] Add mechanism to temporarily replace root bindings of vars

[CLJ-465] with-local-vars broken after changes to make dynamic off by default

[CLJ-462] clojure.core/slurp docstring out of date

[CLJ-460] promote interrupt handling to clojure.repl

[CLJ-458] print-table

[CLJ-454] docstrings for special ops

[CLJ-453] Reflection

[CLJ-448] structural diff

[CLJ-447] biginteger fn misses a case for clojure.lang.BigInt

[CLJ-444] Infinite recursion in Keyword.intern leads to stack overflow

[CLJ-443] Add zero-arity body to comp function

[CLJ-442] Fix embedded doc links on clojure.org docs

[CLJ-441] Add unchecked coercions

[CLJ-439] Automatic type translation from Integer to Long

[CLJ-437] Bugs in clojure.set/subset? and superset? for sets with false/nil elements

[CLJ-435] stackoverflow exception in printing meta with :type

[CLJ-433] munge should not munge $ (which isJavaIdentifierPart), should munge ' (which is not)

[CLJ-752] Remove earmuff == dynamic support

[CLJ-697] Compiler doesn't push a binding for unchecked-math

[CLJ-695] pprint does not respect print-length

[CLJ-305] out has a misleading docstring

[CLJ-286] out being used as java.io.PrintWriter