codeclimate-duplication
is an engine that wraps flay and supports Java, Ruby,
Python, JavaScript, and PHP. You can run it on the command line using the Code
Climate CLI or on our hosted analysis platform.
The duplication engine's algorithm can be surprising, but it's actually very simple. We have a docs page explaining the algorithm.
- Install the Code Climate CLI, if you haven't already.
- You're ready to analyze!
cd
into your project's folder and runcodeclimate analyze
. Duplication analysis is enabled by default, so you don't need to do anything else.
We set useful threshold defaults for the languages we support but you may want to adjust these settings based on your project guidelines.
The mass threshold configuration represents the minimum "mass" a code block must have to be analyzed for duplication. If the engine is too easily reporting duplication, try raising the threshold. If you suspect that the engine isn't catching enough duplication, try lowering the threshold. The best setting tends to differ from language to language.
To adjust this setting, use the top-level checks
key in your config file:
checks:
identical-code:
config:
threshold: 25
similar-code:
config:
threshold: 50
Note that you have the update the YAML structure under the languages
key to
the Hash type to support extra configuration.
By default, the duplication engine will report code that has been duplicated in just two locations. You can be less strict by only raising a warning if code is duplicated in three or more locations only. To adjust this setting, add a count_threshold
key to your config. For instance, to use the default mass_threshold
for ruby, but to enforce the Rule of Three, you could use this configuration:
plugins:
duplication:
enabled: true
config:
languages:
ruby:
count_threshold: 3
You can also change the default count_threshold
for all languages:
plugins:
duplication:
enabled: true
config:
count_threshold: 3
All engines check only appropriate files but you can override default set of
patterns. Patterns are ran against the project root directory so you have to use
**
to match files in nested directories. Also note that you have to specify
all patterns, not only the one you want to add.
plugins:
duplication:
enabled: true
config:
languages:
ruby:
patterns:
- "**/*.rb
- "**/*.rake"
- "Rakefile"
- "**/*.ruby"
By default, the Duplication engine will use a Python 2 parser. To enable
analysis for Python 3 code, specify the python_version
as shown in the example
below. This will enable a Python 3 parser and add the .py3
file extension to
the list of included file patterns.
plugins:
duplication:
enabled: true
config:
languages:
python:
python_version: 3
Sometimes structural similarities are reported that you just don't care about. For example, the contents of arrays or hashes might have similar structures and there's little you can do to refactor them. You can specify language specific filters to ignore any issues that match the pattern. Here is an example that filters simple hashes and arrays:
plugins:
duplication:
enabled: true
config:
languages:
ruby:
filters:
- "(hash (lit _) (str _) ___)"
- "(array (str _) ___)"
The syntax for patterns are pretty simple. In the first pattern:
"(hash (lit _) (str _) ___)"
specifies "A hash with a literal key, a
string value, followed by anything else (including nothing)". You
could also specify "(hash ___)"
to ignore all hashes altogether.
Figuring out what to filter is tricky. codeclimate-duplication comes
with a configuration option to help with the discovery. Instead of
scanning your code and printing out issues for codeclimate, it prints
out the parse-trees instead! Just add dump_ast: true
and debug: true
to your
.codeclimate.yml file:
---
plugins:
duplication:
enabled: true
config:
dump_ast: true
debug: true
... rest of config ...
Then run codeclimate analyze
while using the debug flag to output stderr:
% CODECLIMATE_DEBUG=1 codeclimate analyze
Running that command might output something like:
Sexps for issues:
# 1) ExpressionStatement#4261258897 mass=128:
# 1.1) bogus-examples.js:5
s(:ExpressionStatement,
:expression,
s(:AssignmentExpression,
:"=",
:left,
s(:MemberExpression,
:object,
s(:Identifier, :EventBlock),
:property,
s(:Identifier, :propTypes)),
... LOTS more...)
... even more LOTS more...)
This is the internal representation of the actual code. Assuming you've looked at those issues and have determined them not to be an issue you want to address, you can filter it by writing a pattern string that would match that tree.
Looking at the tree output again, this time flattening it out:
s(:ExpressionStatement, :expression, s(:AssignmentExpression, :"=",:left, ...) ...)
The internal representation (which is ruby) is different from the
pattern language (which is lisp-like), so first we need to convert
s(:
to (
and remove all commas and colons:
(ExpressionStatement expression (AssignmentExpression "=" left ...) ...)
Next, we don't care bout expression
so let's get rid of that by
replacing it with the matcher for any single element _
:
(ExpressionStatement _ (AssignmentExpression "=" left ...) ...)
The same goes for "="
and left
, but we actually don't care about
the rest of the AssignmentExpression node, so let's use the matcher
that'll ignore the remainder of the tree ___
:
(ExpressionStatement _ (AssignmentExpression ___) ...)
And finally, we don't care about what follows in the
ExpressionStatement
so let's ignore the rest too:
(ExpressionStatement _ (AssignmentExpression ___) ___)
This reads: "Any ExpressionStatement node, with any value and an AssignmentExpression node with anything in it, followed by anything else". There are other ways to write a pattern to match this tree, but this is pretty clear.
Then you can add that filter to your config:
---
plugins:
duplication:
enabled: true
config:
dump_ast: true
languages:
javascript:
filters:
- "(ExpressionStatement _ (AssignmentExpression ___) ___)"
Then rerun the analyzer and figure out what the next filter should be.
When you are happy with the results, remove the dump_ast
config (or
set it to false) to go back to normal analysis.
For more information on pattern matching, see sexp_processor, especially sexp.rb