Upgrade compojure and clout deps for compatibility with Clojure 1.7.0-alpha6

[jafingerhut]

See recent compojure and clout commits as of Apr 1 2015 for updates they have made to pull in latest Instaparse (version 1.3.6). Previous versions of Instaparse were not compatible.

[yogthos]

I just updated the dependencies, and I'll push out a new version to clojars tonight

[jafingerhut]

Looks like maybe something else besides upgrading compojure and clout deps is needed for compatibility with Clojure 1.7.0-alpha6, but I don't yet know what it is.

With latest version of lib-noir as of Apr 6 2015, but adding profiles to its project.clj for different versions of Clojure like this:

• :profiles {:1.6 {:dependencies [[org.clojure/clojure "1.6.0"]]}

•         :1.7a5 {:dependencies [[org.clojure/clojure "1.7.0-alpha5"]]}
  

•         :1.7a6 {:dependencies [[org.clojure/clojure "1.7.0-alpha6"]]}}
  

I see this behavior:

lein with-profile +1.6 do clean, test ; no errors
lein with-profile +1.7a5 do clean, test ; no errors
lein with-profile +1.7a6 do clean, test ; errors, an excerpt of them shown below

Exception in thread "main" java.lang.RuntimeException: Unable to resolve symbol: options in this context, compiling:(ring/middleware/format_params.clj:217:28)
at clojure.lang.Compiler.analyze(Compiler.java:6543)
at clojure.lang.Compiler.analyze(Compiler.java:6485)
at clojure.lang.Compiler$InvokeExpr.parse(Compiler.java:3791)

I encountered this error on Mac OS X 10.10.2, Oracle Java 1.8.0_11, Leiningen 2.5.0.
Also Ubuntu Linux 14.04.2, Oracle Java 1.8.0_25, Lein 2.5.0 (also Oracle Java 1.7.0_71)

[yogthos]

I just bumped up ring-middleware-format to the latest version, so hopefully that helps. Otherwise, have to wait until it's compatible with the latest Clojure release.

[MichaelBlume]

I'm gonna bet it's this ngrunwald/ring-middleware-format#37

[jafingerhut]

Thanks, Michael.

From your description of the issue at the link, that jogs my memory that Rich Hickey did commit a change in Clojure 1.7.0-alpha6 such that small literal maps will be PersistentArrayMaps instead of PersistentHashMaps, and I think at the same time he reversed the order that PersistentArrayMaps stored their elements (not sure about that last statement, though -- just a quick guess from looking at the source code change). Here is the commit: clojure/clojure@692645c

(only the last two 'hunks' of that commit are relevant)

[jafingerhut]

With the latest commit to lib-noir on Apr 6 2015, it is no longer throwing an exception. I see 2 tests failing, though. I haven't looked at them closely yet, but it might be due to the tests depending upon the order of key/value pairs in a map, and the change in this I guessed about above in 1.7.0-alpha6.

[yogthos]

yeah the key ordering is the likely culprit

[jafingerhut]

Looks like it might be key ordering, but another possible factor is that deftests are run in an order that can depend upon the seq order of hashes, too. I believe that with Clojure 1.7.0-alpha5, the set-timeout! test is actually running before the set-size! test, because I see a :timeout key in the cache when printing out its contents in the set-size! test with that version of Clojure, but I do not see that key with Clojure 1.7.0-alpha6.

Maybe clear! should be returning the cache back to its initial state of {} rather than only removing the items? Either that or combine some of your deftest contents together so you can control the order they are executed.

[yogthos]

I won't have time to look into this in the near future, but it sounds like your analysis is correct. I'm not opposed to modify clear! behavior and/or changing the flow for the tests though. If you'd like to give it a shot I can push out the update.