An analyzer for host agnostic Clojure code, written in Clojure and producing AST in EDN.
I gave a talk on tools.analyzer[.jvm] at ClojureX in December 2015. Video here
Timothy Baldridge gave a talk on tools.analyzer[.jvm] at Clojure/West in March 2014. Video here.
Note that the analyzer in this library should not to be used directly as it lacks any knowledge about host-specific special forms and it should only be considered as a building platform for host-specific analyzers. Currently the following platform specific analyzers written on top of tools.analyzer exist: tools.analyzer.jvm, tools.analyzer.js
clojure.tools.analyzer/analyze
will not work out of the box, as it requires a number of entry-points to be set.
Here's what could happen trying to use clojure.tools.analyzer/analyze
directly:
clojure.tools.analyzer> (analyze 'a {})
Attempting to call unbound fn: #'clojure.tools.analyzer/macroexpand-1
[Thrown class java.lang.IllegalStateException]
At the moment there exist two official analyzers written on top of tools.analyzer: tools.analyzer.jvm for clojure on the JVM and tools.analyzer.js for clojurescript. We will use tools.analyzer.jvm for those examples.
Here's a simplified version of how clojure.tools.analyzer.jvm/analyze
is defined:
(require '[clojure.tools.analyzer :as ana])
(require '[clojure.tools.analyzer.env :as env])
(defn analyze [form env]
(binding [ana/macroexpand-1 macroexpand-1
ana/create-var create-var
ana/parse parse
ana/var? var?]
(env/ensure (global-env)
(run-passes (-analyze form env))))))
Here, -analyze
is a multimethod that defaults to ana/analyze
and defines analysis methods for the JVM specific special forms, global-env
is a function that returns a global environment for the JVM analyzer and run-passes
is a function that takes an AST and applies a number of passes to it.
The tools.analyzer.jvm
README contains more examples on how the analyze
function works as well as a reference for all the nodes it can return.
One of the most important features of tools.analyzer
is the ability to walk generically through the AST nodes, this has been immensely useful to write most of the passes used by the various analyzers.
The tools.analyzer.ast
namespace provides a number of functions that implement various generic AST walking strategies.
The children
function returns a vector of the children nodes of the current node (the output has been elided and pretty-printed for clarity):
clojure.tools.analyzer.jvm> (require '[clojure.tools.analyzer.ast :as ast])
nil
clojure.tools.analyzer.jvm> (ast/children (analyze '(do 1 2 :foo)))
[{:op :const,
:id 0,
:type :number,
:val 1,
:form 1,
...}
{:op :const,
:id 1,
:type :number,
:val 2,
:form 2,
...}
{:op :const,
:id 3,
:type :keyword,
:val :foo,
:form :foo,
...}]
If we want to access a flattened view of all the nodes of an AST, we can use the nodes
function:
clojure.tools.analyzer.jvm> (ast/nodes (analyze '[1 (+ 1 2)]))
({:op :vector,
:top-level true,
:items
[{:op :const,
:type :number,
:val 1,
...}
{:op :static-call,
:class clojure.lang.Numbers,
:method add,
:form (. clojure.lang.Numbers (add 1 2)),
:args [{:op :const,
:val 1,
...}
{:op :const,
:val 2,
...}],
...}]
:form [1 (+ 1 2)],
...}
{:op :const,
:type :number,
:val 1,
...}
{:op :static-call,
:class clojure.lang.Numbers,
:method add,
:form (. clojure.lang.Numbers (add 1 2)),
:args [{:op :const,
:val 1,
...}
{:op :const,
:val 2,
...}],
...}
..)
The update-children
function takes an AST node and a function and replaces the children nodes of the given node with the result of applying the function to each children node.
clojure.tools.analyzer.jvm> (ast/update-children (analyze '(do 1 (+ 1 2) :foo))
#(assoc % :visited true))
{:op :do
:statements
[{:op :const,
:val 1,
:visited true,
...}
{:op :static-call,
:class clojure.lang.Numbers,
:method add,
:visited true,
:args [{:op :const
:val 1,
...}
{:op :const,
:val 2,
...}],
...}]
:ret
{:op :const,
:val :foo,
:visited true,
...},
...}
If it's desiderable to walk all the AST applying a function to all the nodes and the children nodes, one of walk
, prewalk
or postwalk
should be used, read the docstrings of the three functions to understand the differences.
Here's the previous example using prewalk
instead of update-children
:
clojure.tools.analyzer.jvm> (ast/prewalk (analyze '(do 1 (+ 1 2) :foo))
#(assoc % :visited true))
{:op :do
:visited true,
:statements
[{:op :const,
:val 1,
:visited true,
...}
{:op :static-call,
:class clojure.lang.Numbers,
:method add,
:visited true,
:args [{:op :const
:val 1,
:visited true,
...}
{:op :const,
:val 2,
:visited true,
...}],
...}]
:ret
{:op :const,
:val :foo,
:visited true,
...},
...}
As you can see, this time all the nodes have been marked :visited
.
Since version 0.6.0
, passes can be scheduled automatically using clojure.tools.analyzer.passes/schedule
rather than having to compose them and sort out pass dependencies manually, refer to its docstrings and examples from tools.analyzer.jvm
for more info.
YourKit has given an open source license for their profiler, greatly simplifying the profiling of tools.analyzer performance.
YourKit is kindly supporting open source projects with its full-featured Java Profiler. YourKit, LLC is the creator of innovative and intelligent tools for profiling Java and .NET applications. Take a look at YourKit's leading software products:
Latest stable release: 0.6.9
Leiningen dependency information:
[org.clojure/tools.analyzer "0.6.9"]
Maven dependency information:
<dependency>
<groupId>org.clojure</groupId>
<artifactId>tools.analyzer</artifactId>
<version>0.6.9</version>
</dependency>
Copyright © 2013-2018 Nicola Mometto, Rich Hickey & contributors.
Distributed under the Eclipse Public License, the same as Clojure.
Can you improve this documentation? These fine people already did:
Nicola Mometto, Bronsa, Alex Miller & Andy FingerhutEdit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close