Guide to Writing a Native Bear

Welcome. This document presents information on how to write a bear for coala. It assumes you know how to use coala. If not, please read our main tutorial

The sample sources for this tutorial lie at our coala-tutorial repository, go clone it with:

git clone

All paths and commands given here are meant to be executed from the root directory of the coala-tutorial repository.


If you want to wrap an already existing tool, please refer to this tutorial instead.

What is a bear?

A bear is meant to do some analysis on source code. The source code will be provided by coala so the bear doesn’t have to care where it comes from or where it goes.

There are two kinds of bears:

  • LocalBears, which only perform analysis on each file itself
  • GlobalBears, which are project wide, like the GitCommitBear

A bear can communicate with the user via two ways:

  • Via log messages
  • Via results

Log messages will be logged according to the users settings and are usually used if something goes wrong. However you can use debug for providing development related debug information since it will not be shown to the user by default. If error/failure messages are used, the bear is expected not to continue analysis.

A Hello World Bear

Below is the code given for a simple bear that sends a debug message for each file:

from coalib.bears.LocalBear import LocalBear

class HelloWorldBear(LocalBear):
    def run(self,
        self.debug("Hello World! Checking file", filename, ".")

This bear is stored at ./bears/

In order to let coala execute this bear you need to let coala know where to find it. We can do that with the -d (--bear-dirs) argument:

coala -f src/*.c -d bears -b HelloWorldBear -L DEBUG --flush-cache


The given bear directories must not have any glob expressions in them. Any character that could be interpreted as a part of a glob expression will be escaped. Please use comma separated values to give several such directories instead. Do not forget to flush the cache (by adding the argument --flush-cache when running coala) if you run a new bear on a file which has been previously analyzed (by coala).

You should now see an output like this on your command line:

[WARNING][15:07:39] Default coafile '.coafile' not found!
Here's what you can do:
* add `--save` to generate a config file with your current options
* add `-I` to suppress any use of config files
[DEBUG][15:07:39] Platform Linux -- Python 3.5.2, coalib
[DEBUG][15:07:39] The file cache was successfully flushed.
[DEBUG][15:07:39] Files that will be checked:
[DEBUG][15:07:40] coala is run only on changed files, bears' log
messages from previous runs may not appear. You may use the
`--flush-cache` flag to see them.
[DEBUG][15:07:40] Running bear HelloWorldBear...
[DEBUG][15:07:40] Hello World! Checking file /home/LordVoldemort/
programs/coa_dir/coala-tutorial/src/main.c .

Notice that the last ``[DEBUG]`` message is what was coded in
````. All the other messages are inherited from the
``LocalBear`` class or run by the code responsible for executing the


The first WARNING message is because our directory, does not contain a .coafile. If you have followed the instructions in our main tutorial, you will have a .coafile in your working directory. Its best if you delete that file before working on this tutorial, else you will see a bunch of other outputs from other bears as well.

The Bear class also supports warn and err to create WARNING and ERROR messages respectively.

Communicating with the User

Now we can send messages through the queue, we can do the real work. Let’s say:

  • We want some information from the user (e.g. the tab width if we rely on indentation).
  • We’ve got some useful information for the user and want to show it to them. This might be some issue with their code or just an information like the number of lines.

So let’s extend our HelloWorldBear a bit, I’ve named the new bear with the creative name CommunicationBear:

from coalib.bears.LocalBear import LocalBear

class CommunicationBear(LocalBear):

    def run(self,
            user_input: str):
        Communicates with the user.

        :param user_input: Arbitrary user input.
        self.debug("Got '{ui}' as user input of type {type}.".format(

        yield self.new_result(message="A hello world result.",

Try executing it:

coala -f=src/\*.c -d=bears -b=CommunicationBear -L=DEBUG --flush-cache

Hey, we’ll get asked for the user_input!

[WARNING][15:20:18] Default coafile '.coafile' not found!
Here's what you can do:
* add `--save` to generate a config file with your current options
* add `-I` to suppress any use of config files
Please enter a value for the setting "user_input" (No description given.)
needed by CommunicationBear for section "cli":

Wasn’t that easy? Go ahead, enter something and observe the output.

Avada Kedavra
[DEBUG][15:22:55] Platform Linux -- Python 3.5.2, coalib
[DEBUG][15:22:55] The file cache was successfully flushed.
[DEBUG][15:22:55] Files that will be checked:
[DEBUG][15:22:55] coala is run only on changed files, bears' log messages
from previous runs may not appear. You may use the `--flush-cache` flag to
see them.
[DEBUG][15:22:55] Running bear CommunicationBear...
[DEBUG][15:22:55] Got 'Avada Kedavra' as user input of type <class 'str'>.

**** CommunicationBear [Section: cli] ****

!    ! [Severity: NORMAL]
!    ! A hello world result.
[    ] Do (N)othing
[    ] (O)pen file
[    ] Add (I)gnore comment
[    ] Enter number (Ctrl-D to exit):

So, what did coala do here?

First, coala looked at the parameters of the run method and found that we need some value named user_input. Then it parsed our documentation comment and found a description for the parameter which was shown to us to help us choose the right value. After the needed values are provided, coala converts us the value into a string because we’ve provided the str annotation for this parameter. If no annotation is given or the value isn’t convertible into the desired data type, you will get a coalib.settings.Setting.Setting.

Your docstring can also be used to tell the user what exactly your bear does.

Try executing

coala -d bears -b CommunicationBear --show-bears --show-description

This will show the user a bunch of information related to the bear like: - A description of what the bear does - The sections which uses it - The settings it uses (optional and required)


The bears are not yet installed. We still have to specify the bear directory using -d or --bear-dirs flag.

Install locally Written Bears

Let’s say that we wrote a file that contain our NewBear and we want to run it locally. To install our NewBear:

  • Move the to our clone of coala-bears in coala-bear/bears/<some_directory>.
  • Update all bears from source with:
pip install -U <path/to/coala-bears>

Our NewBear is installed.

Try Executing:

coala --show-bears

This shows a list of all installed bears. We can find our NewBear in the list.

What Data Types are Supported?

The Setting does support some very basic types:

  • String (str)
  • Float (float)
  • Int (int)
  • Boolean (bool, will accept values like true, yes, yeah, no, nope, false)
  • List of strings (list, values will be split by comma)
  • Dict of strings (dict, values will be split by comma and colon)

If you need another type, you can write the conversion function yourself and use this function as the annotation (if you cannot convert value, be sure to throw TypeError or ValueError). We’ve provided a few advanced conversions for you:

  • coalib.settings.Setting.path, converts to an absolute file path relative to the file/command where the setting was set
  • coalib.settings.Setting.path_list, converts to a list of absolute file paths relative to the file/command where the setting was set
  • coalib.settings.Setting.typed_list(typ), converts to a list and applies the given conversion (typ) to each element.
  • coalib.settings.Setting.typed_ordered_dict(key_type, value_type, default), converts to a dict while applying the key_type conversion to all keys, the value_type conversion to all values and uses the default value for all unset keys. Use typed_dict if the order is irrelevant for you.


In the end we’ve got a result. If a file is provided, coala will show the file, if a line is provided, coala will also show a few lines before the affecting line. There are a few parameters to the Result constructor, so you can e.g. create a result that proposes a code change to the user. If the user likes it, coala will apply it automatically - you don’t need to care.

Your function needs to return an iterable of Result objects: that means you can either return a list of Result objects or simply yield them and write the method as a generator.


We are currently planning to simplify Bears for bear writers and us. In order to make your Bear future proof, we recommend writing your method in generator style.

Don’t worry: in order to migrate your Bears to our new API, you will likely only need to change two lines of code. For more information about how bears will look in the future, please read up on or ask us on

Bears Depending on Other Bears

So we’ve got a result, but what if we need our Bear to depend on results from a different Bear?

Well coala has an efficient dependency management system that would run the other Bear before your Bear and get its results for you. All you need to do is to tell coala which Bear(s) you want to run before your Bear.

So let’s see how you could tell coala which Bears to run before yours:

from coalib.bears.LocalBear import LocalBear
from bears.somePathTo.OtherBear import OtherBear

class DependentBear(LocalBear):

    BEAR_DEPS = {OtherBear}

    def run(self, filename, file, dependency_results):
        results = dependency_results[]

As you can see we have a BEAR_DEPS set which contains a list of bears we wish to depend on. In this case it is a set with 1 item: “OtherBear”.


The BEAR_DEPS set must have classes of the bear itself, not the name as a string.

coala gets the BEAR_DEPS before executing the DependentBear and runs all the Bears in there first.

After running these bears, coala gives all the results returned by the Bears in the dependency_results dictionary, which has the Bear’s name as a key and a list of results as the value. E.g. in this case, we would have dependency_results == {'OtherBear' : [list containing results of OtherBear]]}.


dependency_results is a keyword here and it cannot be called by any other name.

Hidden Results

Apart from regular Results, coala provides HiddenResults, which are used to share data between Bears as well as giving results which are not shown to the user. This feature is specifically for Bears that are dependencies of other Bears, and do not want to return Results which would be displayed when the bear is run.

Let’s see how we can use HiddenResults in our Bear:

from coalib.bears.LocalBear import LocalBear
from coalib.results.HiddenResult import HiddenResult

class OtherBear(LocalBear):

    def run(self, filename, file):
        yield HiddenResult(self, ["Some Content", "Some Other Content"])

Here we see that this Bear (unlike normal Bears) yields a HiddenResult instead of a Result. The first parameter in HiddenResult should be the instance of the Bear that yields this result (in this case self), and second argument should be the content we want to transfer between the Bears. Here we use a list of strings as content but it can be any object.

More Configuration Options

coala provides metadata to further configure your bear according to your needs. Here is the list of all the metadata you can supply:


To indicate which languages your bear supports, you need to give it a set of strings as a value:

class SomeBear(Bear):
    LANGUAGES = {'C', 'CPP','C#', 'D'}


To indicate the requirements of the bear, assign REQUIREMENTS a set with instances of subclass of PackageRequirement such as:

  • PipRequirement
  • NpmRequirement
  • CondaRequirement
  • DistributionRequirement
  • GemRequirement
  • GoRequirement
  • JuliaRequirement
  • RscriptRequirement
class SomeBear(Bear):
    PipRequirement('coala_decorators', '0.2.1')}

To specify multiple requirements you can use the multiple method. This can receive both tuples of strings, in case you want a specific version, or a simple string, in case you want the latest version to be specified.

class SomeBear(Bear):
    REQUIREMENTS = PipRequirement.multiple(
        ('colorama', '0.1'),


If your bear needs to include local files, then specify it by giving strings containing file paths, relative to the file containing the bear, to the INCLUDE_LOCAL_FILES.

class SomeBear(Bear):
    INCLUDE_LOCAL_FILES = {'checkstyle.jar',


To easily keep track of what a bear can do, you can set the value of CAN_FIX and CAN_DETECT sets.

class SomeBear(Bear):
    CAN_DETECT = {'Unused Code', 'Spelling'}

    CAN_FIX = {'Syntax', 'Formatting'}

To view a full list of possible values, check this list:

  • Syntax
  • Formatting
  • Security
  • Complexity
  • Smell
  • Unused Code
  • Redundancy
  • Variable Misuse
  • Spelling
  • Memory Leak
  • Documentation
  • Duplication
  • Commented Code
  • Grammar
  • Missing Import
  • Unreachable Code
  • Undefined Element
  • Code Simplification

Specifying something to CAN_FIX makes it obvious that it can be detected too, so it may be omitted from CAN_DETECT


BEAR_DEPS contains bear classes that are to be executed before this bear gets executed. The results of these bears will then be passed to the run method as a dict via the dependency_results argument. The dict will have the name of the Bear as key and the list of its results as value:

class SomeOtherBear(Bear):
    BEAR_DEPS = {SomeBear}

For more detail see Bears Depending on Other Bears.

Other Metadata


class SomeBear(Bear):
    AUTHORS = {'Jon Snow'}
    AUTHORS_EMAILS = {'[email protected]'}
    MAINTAINERS = {'Catelyn Stark'}
    MAINTAINERS_EMAILS = {'[email protected]'}
    LICENSE = 'AGPL-3.0'
    SEE_MORE = ''