/$$
| $$
/$$$$$$$ | $$ /$$ /$$ /$$
/$$_____/ | $$ |__/ | $$ /$$/
| $$ | $$ /$$ \ $$$$/
| $$ | $$ | $$ >$$ $$
| $$$$$$$ | $$ | $$ /$$/\ $$
\_______/ |__/ | $$ |__/ \__/
/$$ | $$
| $$$$$$/
\______/
This plugin is deprecated
After three years of auto-generating files on disk, cluttering up project.clj
configurations, and otherwise being a useful but ugly hack, I am happy to announce that the reign of cljx is over: Clojure 1.7 brings us glorious Reader Conditionals (AKA .cljc
files).
Reader Conditionals provide richer semantics than cljx like default expressions and form splicing.
As an addition to the core language, Reader Conditionals have the vastly improved user experience of "no build tool configuration necessary".
I encourage all projects that can use Clojure 1.7 or higher to use Reader Conditionals rather than cljx.
If you need reader conditional support for projects using Clojure 1.6 or earlier, see the cljsee Leiningen plugin.
Chas and I will not be doing additional work on cljx, but we are available for commerical support if you absolutely need it.
For more details on cross-platform code rewriting and the historical context leading up to Reader Conditionals, see:
- Reader Conditional design doc: http://dev.clojure.org/display/design/Reader+Conditionals
- "Feature expressions - just because?" clojure-dev thread: https://groups.google.com/d/msg/clojure-dev/6pnIeXFRwnI/AGOQAwA2vrcJ
- Alan Dipert's + Micha Niskin's "Feature Macros": https://groups.google.com/d/msg/clojure-dev/LW0ocQ1RcYI/PDOpmTBt-VIJ
README for deprecated plugin follows
cljx is a Leiningen plugin and nREPL middleware that produces Clojure and ClojureScript code from a single annotated codebase. Effectively, it is an s-expression preprocessor that yields either Clojure and ClojureScript sources on disk (e.g. for inclusion in jars or for input to compilation tools like Clojure AOT-compilation and lein-cljsbuild):
+---------+
| .cljx |
| sources |
+-----+---+
|
|
|
+-----v-----+ +----------------+
| cljx <-------+ configuration |
| Leiningen | | + |
| plugin | | rules |
+--+--+-----+ +----------------+
| |
| |
+------------+ | | +------------+
| .clj <-+ +-> .cljs |
| sources | | sources |
+------------+ +------------+
…or, when used in a REPL, cljx automatically applies the same transformation to
any namespaces to be loaded (e.g. as a result of :require
declarations) from
.cljx
files before they are consumed by the Clojure or ClojureScript compiler:
+------------+ +---------------+
+---------+ | cljx | | configuration |
| .cljx | | nREPL | | + |
| sources +----------> middleware <-------+ rules |
+---------+ +-----+------+ +---------------+
|
|
|
+-------v---------+
| `require` |
| `:require` |
+---------+ | `load-namespace`| +---------+
| .cljs +-------> ...etc... <--------+ .cljs |
| sources | +------+--+-------+ | sources |
+---------+ | | +---------+
| |
+------------+ | | +---------------+
| Clojure <-+ +-> ClojureScript |
| compiler | | compiler |
+------------+ +---------------+
When using cljx, you put APIs and implementations that are meant to be
fundamentally portable between Clojure and ClojureScript into one annotated
.cljx
codebase, and leave things that are necessarily tied to a single
compilation target in their "native" language (e.g. macros should always be in
Clojure sources, DOM manipulation stuffs always in ClojureScript sources, etc).
Does this seem crazy? Crazy awesome, maybe.
Projects that use cljx
Here's some real-world examples of projects that use cljx (when you feel in trouble, refer to these for usage and configuration examples):
- schema
- formative
- pprng
- sablono
- double-check
- garden
- hickory
- validateur
- frak
- cellular
- enoki
- inflections-clj
- pathetic
- cats
"Installation"
To use it, add to the :dev
profile in your project.clj
:
:profiles {:dev {:plugins [[com.keminglabs/cljx "0.6.0"]]}}
:cljx {:builds [{:source-paths ["src/cljx"]
:output-path "target/classes"
:rules :clj}
{:source-paths ["src/cljx"]
:output-path "target/classes"
:rules :cljs}]}
A more comprehensive configuration example can be found here.
If you want cljx to be invoked automatically as part of the Leiningen
compilation process (e.g. before cutting a jar, performing a release), add cljx
to your :prep-tasks
vector in project.clj
. The :prep-tasks project value
will replace the default leiningen setting (below) which is required for AOT
compilation:
:prep-tasks ["javac" "compile"]
With cljx once:
:prep-tasks [["cljx" "once"] "javac" "compile"]
Note that any Clojure tests generated by cljx will not be picked up by Leiningen's test runner (see this for some background). In general, it is probably advisable to set up aliases so you can completely control what Leiningen will run:
:aliases {"cleantest" ["do" "clean," "cljx" "once," "test," "cljsbuild" "test"]}
Changelog
See CHANGES.md
at the root of this repo.
(You'll especially want to look at the entry for 0.3.0
if you've been using
older versions of cljx, as things have changed [of course, we think
significantly for the better :-P].)
Usage
There are two ways in which the cljx transformation can be made: via a Leiningen task (necessary when you need the transformation result on disk for e.g. packaging into a jar for distribution), and/or via an nREPL middleware that makes using the Leiningen task unnecessary in REPL sessions.
cljx's Leiningen task can be run once
or auto
; if the latter (e.g. lein cljx auto
), it will watch all source-paths
for changes to .cljx
files.
once
is the default.
Each build (i.e. maps in the :builds
vector in the :cljx
configuration) can
be configured with the following options:
:source-paths
, a sequence of the source roots that contain your.cljx
files. Note that putting your.cljx
files in your "regular" Leiningen project's:source-paths
(by default,"src"
) is not recommended; doing so will likely lead to them being included in e.g. jar files created by Leiningen. Better to keep them separate, and use cljx to direct Clojure and ClojureScript sources whereever they will be picked up by other tooling.:output-path
, the root directory where cljx's output will land. Common options are"target/classes"
for both Clojure and ClojureScript files you plan on distributing as a library; or, in an application project using lein-cljsbuild to produce deployable JavaScript, sending cljx-produced Clojure output to"target/classes"
(so it's on the classpath and available to be added to a jar/war) and ClojureScript output to a dummy directory (e.g."target/generated/cljs"
) that can be a source path in your lein-cljsbuild configuration(s).:rules
can be one of::clj
or:cljs
to use cljx's default Clojure or ClojureScript ruleset (cljx.rules/clj-rules
andcljx.rules/cljs-rules
, respectively)- a map that specifies the three slots that make up a cljx ruleset:
:filetype
, a string that defines what the extension of output filenames will be, e.g."cljs"
:features
, a set of strings, each naming an enabled "feature"; code in.cljx
files that is annotated with a feature that is not included in this set will be pruned in the output:transforms
, a sequence of functions that are applied to each expression in each input file, and can modify that expression without constraint
- a fully-qualified symbol that names a var containing a map as described above
In general, you'll never need to go beyond the named cljx-provided rules.
E.g., the .cljx
source containing
(ns example
(#+clj :use #+cljs :use-macros [c2.macros :only (combine-with)]))
(defn x-to-string
[x]
(let [buf #+clj (StringBuilder.) #+cljs (gstring/StringBuffer.)]
(.append buf "x is: ")
(.append buf (str x))))
(reify
#+clj clojure.lang.IFn
#+cljs cljs.core.IFn
(invoke [_ x] (inc x)))
…will, when transformed using the :cljs
ruleset, yield:
(ns example
( :use-macros [c2.macros :only (combine-with)]))
(defn x-to-string
[x]
(let [buf (gstring/StringBuffer.)]
(.append buf "x is: ")
(.append buf (str x))))
(reify
cljs.core.IFn
(invoke [_ x] (inc x)))
Notice that only the #+cljs
-annotated expressions remain, and that everything
is still in the same position as it was in the .cljx
file; this last
fact means that line and column numbers produced by the resulting
Clojure/ClojureScript code (e.g. in error messages, stack traces/frames,
debuggers, source maps, etc) will remain true to the original sources.
The #+feature-name
"annotation" syntax is shamelessly stolen from Common
Lisp (and is
perhaps being considered for inclusion in Clojure[Script] itself?...see feature
expressions). cljx
only supports the simplest form of the syntax; other forms can be considered
valid TODOs:
- Exclusionary annotations, e.g.
#-cljs
- "Union" annotations, e.g.
#+(or clj clr)
Clojure is a hosted language, in all flavours
cljx does not try to hide implementation differences between host platforms. Clojure has ints, floats, longs, &c., ClojureScript has number; Clojure regular expressions act differently than ClojureScript regular expressions, because they are different, and so on.
cljx only tries to unify Clojure/ClojureScript abstractions when it makes sense.
E.g., converting clojure.lang.IFn
into IFn
when generating ClojureScript.
The rest is up to you, in annotating your code to include or exclude what's
needed by each runtime.
Also, note that cljx has no effect on code produced by macros. Macroexpansion occurs long after cljx touches your code.
REPL Integration
cljx provides an nREPL middleware that allows you to work with .cljx
files in
the same way you work with regular .clj
files from any toolchain with good
nREPL support, like cider,
Counterclockwise, etc.
When you add cljx as a :plugin
to your Leiningen project:
- The cljx and Piggieback nREPL
middlewares will automatically be added to your
:repl-options
- cljx itself will be added as a project dependency (this will only affect REPL
processes, and won't leak out into your project's
pom.xml
(as long as you add cljx to your:dev
profile), influencing downstream users of your library, if you're writing one)
(Note that this does not conflict with using the Austin plugin to automate the configuration of your project to use Piggieback. In fact, the pairing is highly recommended for making the ClojureScript REPL side of your cljx project easy-peasy.)
With cljx installed as a plugin, all nREPL evaluations and load-file
operations will be processed by cljx appropriately before they reach the Clojure
or ClojureScript compiler. Whether cljx code is processed for Clojure or
ClojureScript is determined by the existence [or not] of a Piggieback
ClojureScript environment in your current nREPL session's environment; this is
entirely automatic.
Currently, only cljx's default rulesets are used in this case (though you can work around this by making your own higher-order cljx nREPL middleware that uses whatever rulesets you want).
Misc
Syntax highlighting
Get the same syntax highlighting of .cljx
files as you currently do for .clj
files!
Emacs
clojure-mode
supports .cljx
source files out-of-the-box. It will even font-lock features properly.
Vim
autocmd BufNewFile,BufReadPost *.cljx setfiletype clojure
Eclipse + CounterClockwise
- In Preferences, go to General > Editors > File Associations.
- Add a
*.cljx
file type in the upper list. - Add an editor association for that
*.cljx
file type to Counterclockwise'sClojure Editor
.
Thanks
- @jonase and @ohpauleez for enabling the first kibit-based cljx
- @cemerick for design chats, maintaining and extending cljx, and rewriting the core to use sjacket
- @cgrand and @trptcolin for sjacket
- @swannodette for core.match