Humbug is a Mutation Testing framework for PHP. It is currently in development and so, while it does actually work quite well, it will have rough edges that a team of minions are slaving to hammer out. If it falls out of the gate, you have been warned ;).
$ git remote set-url upstream https://github.com/humbug/humbug.gitReplace upstream with the name of the remote you use locally; upstream is commonly
used but you may be using something else. You may also using a different URL (e.g. [email protected]:mockery/mockery.git).
Run git remote -v to see what you're actually using.
Table of Contents
Mutation Testing is, in a nutshell, giving your unit tests a run for their money. It involves injecting small defects into source code and then checking if the unit tests noticed. If they do, then your unit tests have "killed" the mutation. If not, the mutation has escaped detection. As unit tests are intended to prevent regressions, having a real regression pass unnoticed would be a bad thing!
Whereas Code Coverage can tell you what code your tests are executing, Mutation Testing is intended to help you judge how well your unit tests actually perform and where they could be improved.
I've written in more detail about why Mutation Testing is worth having: Lies, Damned Lies and Code Coverage: Towards Mutation Testing
You can clone and install Humbug's dependencies using Composer:
git clone https://github.com/humbug/humbug.git
cd humbug
/path/to/composer.phar installThe humbug command is now at bin/humbug.
If you don't want to track the master branch directly, you can install the Humbug phar as follows:
wget https://padraic.github.io/humbug/downloads/humbug.phar
wget https://padraic.github.io/humbug/downloads/humbug.phar.pubkey
# If you wish to make humbug.phar directly executable
chmod +x humbug.pharOn Windows, you can just download using a browser or from Powershell v3 using the
following commands where wget is an alias for Invoke-WebRequest:
wget https://padraic.github.io/humbug/downloads/humbug.phar -OutFile humbug.phar
wget https://padraic.github.io/humbug/downloads/humbug.phar.pubkey -OutFile humbug.phar.pubkeyIf you're stuck with Powershell v2:
$client = new-object System.Net.WebClient
$client.DownloadFile("https://padraic.github.io/humbug/downloads/humbug.phar", "humbug.phar")
$client.DownloadFile("https://padraic.github.io/humbug/downloads/humbug.phar.pubkey", "humbug.phar.pubkey")The phar is signed with an openssl private key. You will need the pubkey file
to be stored beside the phar file at all times in order to use it. If you rename
humbug.phar to humbug, for example, then also rename the key from
humbug.phar.pubkey to humbug.pubkey.
The phar releases are currently done manually so they will not be updated with the same frequency as git master. To update your current phar, just run:
./humbug.phar self-updateNote: Using a phar means that fixes may take longer to reach your version, but there's more assurance of having a stable development version. The public key is downloaded only once. It is re-used by self-update to verify future phar releases.
Once releases commence towards stable, there will be an alpha, beta, RC and a final release. Your development track phar file will self-update automatically until it reaches a stable release. If you wish to continue tracking the development level phars, you will need to indicate this using one of the stability flags:
./humbug.phar self-update --devIf you experience any issues self-updating with unexpected openssl or SSL errors,
please ensure that you have enabled the openssl extension. On Windows, you can
do this by adding or uncommenting the following line in the php.ini file for
PHP on the command line (if different than the file for your http server):
extension=php_openssl.dll
Certain other SSL errors may arise due missing certificates. You can rectify this
by finding their location on your system (e.g. C:/xampp/php/ext/cacert.pem), or
alternatively downloading a copy from http://curl.haxx.se/ca/cacert.pem. Then
ensure the following option is correctly pointing to this file:
openssl.cafile=C:/path/to/cacert.pem
Due to Humbug's dependencies being pegged to recent versions, adding Humbug to composer.json may give rise to conflicts. The above two methods of installation are preferred where this occurs. You can however install it globally as any other general purpose tool:
composer global require 'humbug/humbug=~1.0@dev'And if you haven't done so previously...add this to ~/.bash_profile (or ~/.bashrc):
export PATH=~/.composer/vendor/bin:$PATHHumbug currently works on PHP 5.4 or greater.
Humbug is still under development so, to repeat, beware of rough edges.
To configure humbug in your project you may run:
humbug configureThis tool will ask some questions required to create the Humbug configuration file
(humbug.json.dist).
In the base directory of your project create a humbug.json.dist file:
{
"timeout": 10,
"source": {
"directories": [
"src"
]
},
"logs": {
"text": "humbuglog.txt",
"json": "humbuglog.json"
}
}You can commit the humbug.json.dist to your VCS and override it locally with a
humbug.json file.
Edit as appropriate. If you do not define at least one log, detailed information about escaped mutants will not be available. The Text log is human readable. If source files exist in the base directory, or files in the source directories must be excluded, you can add exclude patterns (here's one for files in base directory where composer vendor and Tests directories are excluded):
{
"timeout": 10,
"source": {
"directories": [
"."
],
"excludes": [
"vendor",
"Tests"
]
},
"logs": {
"text": "humbuglog.txt"
}
}If, from your project's base directory, you must run tests from another directory then you can signal this also. You should not need to run Humbug from any directory other than your project's base directory.
{
"chdir": "tests",
"timeout": 10,
"source": {
"directories": [
"src"
],
}
}Ensure that your tests are all in a passing state (incomplete and skipped tests are allowed). Humbug will quit if any of your tests are failing.
The magic command, while in your project's base directory (using the PHAR download) is:
./humbug.pharor if you just cloned Humbug:
../humbug/bin/humbugIf all went well, you will get something similar to:
_ _ _
| || |_ _ _ __ | |__ _ _ __ _
| __ | || | ' \| '_ \ || / _` |
|_||_|\_,_|_|_|_|_.__/\_,_\__, |
|___/
Humbug version 1.0-dev
Humbug running test suite to generate logs and code coverage data...
361 [==========================================================] 28 secs
Humbug has completed the initial test run successfully.
Tests: 361 Line Coverage: 64.86%
Humbug is analysing source files...
Mutation Testing is commencing on 78 files...
(.: killed, M: escaped, S: uncovered, E: fatal error, T: timed out)
.....M.M..EMMMMMSSSSMMMMMSMMMMMSSSE.ESSSSSSSSSSSSSSSSSM..M.. | 60 ( 7/78)
...MM.ES..SSSSSSSSSS...MMM.MEMME.SSSS.............SSMMSSSSM. | 120 (12/78)
M.M.M...TT.M...T.MM....S.....SSS..M..SMMSM...........M...... | 180 (17/78)
MM...M...ESSSEM..MMM.M.MM...SSS.SS.M.SMMMMMMM..SMMMMS....... | 240 (24/78)
.........SMMMSMMMM.MM..M.SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS | 300 (26/78)
SSSSSSSSM..E....S......SS......M.SS..S..M...SSSSSSSS....MMM. | 360 (37/78)
.M....MM..SM..S..SSSSSSSS.EM.S.E.M............M.....M.SM.M.M | 420 (45/78)
..M....MMS...MMSSS................M.....EME....SEMS...SSSSSS | 480 (52/78)
SSSSS.EMSSSSM..M.MMMM...SSE.....MMM.M..MM..MSSSSSSSSSSSSSSSS | 540 (60/78)
SSS....SSSSSSSSMM.SSS..........S..M..MSSMS.SSSSSSSSSSSSSSSSS | 600 (68/78)
......E...M..........SM.....M..MMMMM.MMMMMSSSSSSSM.SS
653 mutations were generated:
284 mutants were killed
218 mutants were not covered by tests
131 covered mutants were not detected
17 fatal errors were encountered
3 time outs were encountered
Metrics:
Mutation Score Indicator (MSI): 47%
Mutation Code Coverage: 67%
Covered Code MSI: 70%
Remember that some mutants will inevitably be harmless (i.e. false positives).
Humbug results are being logged as JSON to: log.json
Humbug results are being logged as TEXT to: log.txt
To explain the perhaps cryptic progress output:
- Killed Mutation (.): A mutation that caused unit tests to fail which is a positive outcome.
- Escaped Mutation (M): A mutation where the unit tests still passed which is not what we want! Our unit tests should detect any behaviour changes.
- Uncovered Mutation (S): A mutation which occurs on a line not covered by any unit test. Since there are no unit tests, this is another undesireable result.
- Fatal Error (E): A mutation created a fatal error. Usually a positive result since it's obviously going to be noticed. In some cases, however, it might be a Humbug problem that needs fixing.
- Timeout (T): This is where unit tests exceed the allowed timeout configured for Humbug. Likely a positive result if your timeout is appropriate, and often occurs when a mutation ends up creating an infinite loop.
Kills, errors and timeouts are all counted as detected mutations. We report errors in the logs on the off chance that Humbug itself encountered an internal error, i.e. a bug to be reported as an issue here!
The example summary results reported a number of metric scores:
- A Mutation Score Indicator (MSI) of 47%. This means that 47% of all generated mutations were detected (i.e. kills, timeouts or fatal errors). The MSI is the primary Mutation Testing metric. Given the code coverage of 65%, there is a 18% discrepancy so Code Coverage was a terrible quality measurement in this example.
- Mutation Code Coverage is 67%. On average it should be within the same ballpark as your normal code coverage, but code coverage ignores mutation frequency.
- The Mutation Score Indicator for code that is actually covered by tests was 70% (i.e. ignoring code not even tested). This gives you some idea of how effective the tests that do exist really are.
If you examine these metrics, the standout issue is that the MSI of 47% is 18 points lower than the reported Code Coverage at 65%. These unit tests are far less effective than Code Coverage alone could detect.
Interpreting these results requires some context. The logs will list all undetected mutations as diffs against the original source code. Examining these will provide further insight as to what specific mutations went undetected.
Humbug has a few command line options of note, other than those normally associated with any Symfony Console application.
You can manually set the timeout threshold for any single test:
humbug --timeout=10If you're only interested in mutating a subset of your files, you can pass
any number of --file options containing simple file names, globs or regular
expressions. Basically, these are all passed to the Symfony Finder's name()
method.
humbug --file=NewClass.php --file=*Driver.phpThis in no way restricts the initial Humbug check on the overall test suite which is still executed in full to ensure all tests are passing correctly before proceeding.
Incremental Analysis (IA) is an experimental unfinished mode of operation where results are cached locally between runs and reused where it makes sense. At present, this mode operates very naively by eliminating test runs where both the immediate file being mutated and the relevant tests for a mutated line have not been modified since the last run (as determined by comparing the SHA1 of the files involved).
humbug --incrementalThe IA mode offers a significant performance increase for relatively stable code bases, and you're free to test it and see how it fares in real life. In the future, it does need to take into accounts changes in files which contain parent classes, imported traits and the classes of its immediate dependencies, all of which have an impact on the behaviour of any given object.
IA utilises a local permanent cache, e.g. /home/padraic/.humbug.
Mutation Testing has traditionally been slow. The concept being to re-run your test suite for each mutation generated. To speed things up significantly, Humbug does the following:
- On each test run, it only uses those test classes which cover the specific file and line on which the mutation was inserted.
- It orders test classes to run so that the slowest go last (hopefully the faster tests will detect mutations early!).
- If a mutation falls on a line not covered by any tests, well, we don't bother running any tests.
- Performance may, depending on the source code, be significantly impacted by timeouts. The default of 60s may be far too high for smaller codebases, and far too low for larger ones. As a rule of thumb, it shouldn't exceed the seconds needed to normally run the tests being mutated (and can be set lower).
While all of this speeds up Humbug, do be aware that a Humbug run will be slower than unit testing. A 2 second test suite may require 30 seconds for mutation testing. Or 5 minutes. It all depends on the interplay between lines of code, number of tests, level of code coverage, and the performance of both code and tests.
Humbug implements a basic suite of Mutators, which essentially tells us when a particular PHP token can be mutated, and also apply that mutation to an array of tokens.
Binary Arithmetic:
| Original | Mutated | Original | Mutated |
|---|---|---|---|
| + | - | /= | *= |
| - | + | %= | *= |
| * | / | **= | /= |
| / | * | & | | |
| % | * | | | & |
| ** | / | ^ | & |
| += | -= | ~ | |
| -= | += | >> | << |
| *= | /= | << | >> |
Boolean Substitution:
This temporarily encompasses logical mutators.
| Original | Mutated |
|---|---|
| true | false |
| false | true |
| && | || |
| || | && |
| and | or |
| or | and |
| ! |
Conditional Boundaries:
| Original | Mutated |
|---|---|
| > | >= |
| < | <= |
| >= | > |
| <= | < |
Negated Conditionals:
| Original | Mutated | Original | Mutated |
|---|---|---|---|
| == | != | > | <= |
| != | == | < | >= |
| <> | == | >= | < |
| === | !== | <= | > |
| !== | === |
Increments:
| Original | Mutated |
|---|---|
| ++ | -- |
| -- | ++ |
Return Values:
| Original | Mutated | Original | Mutated |
|---|---|---|---|
| return true; | return false; | return 1.0>; | return -( + 1); |
| return false; | return true; | return $this; | return null; |
| return 0; | return 1; | return function(); | function(); return null; |
| return ; | return 0; | return new Class; | new Class; return null; |
| return 0.0; | return 1.0; | return (Anything); |
(Anything); return null; |
| return 1.0; | return 0.0; |
Literal Numbers:
| Original | Mutated |
|---|---|
| 0 | 1 |
| 1 | 0 |
| Int > 1 | Int + 1 |
| Float >= 1 / <= 2 | Float + 1 |
| Float > 2 | 1 |
If Statements:
All if statements are covered largely by previous mutators, but there are special
cases such as using native functions or class methods without any compares or operations, e.g.
is_int() or in_array(). This would not cover functions defined in files since
they don't exist until runtime (something else to work on!).
| Original | Mutated |
|---|---|
| if(is_int(1)) | if(!is_int(1)) |
More Mutators will be added over time.
bin/humbug stats ../my-project/humbuglog.json ../my-project/list-of-classes.txt --skip-killed=yes [-vvv]Parses stats from humbuglog.json or your custom named JSON log.
CLI reference:
humbug stats [humbuglog.json location] [class list location] [--skip-killed=yes] [-vvv]humbuglog.json location, defaults to ./humbuglog.json
class list location, a path to a text file containing full class names, one per line.
only this files-related stats would be shown
--skip-killed=yes is used to completely skip output of "killed" section
various verbosity levels define amount of info to be displayed:
by default, there's one line per class with amount of mutants killed/escaped/errored/timed out (depending on output section)
-v adds one line per each mutant with line number and method name
-vv adds extra line for each mutant, displaying diff view of line mutant is detected in
-vvv shows full diff with several lines before and after
This can be tested on humbug itself, by running in humbug's dir:
bin/humbug bin/humbug stats [-vvv]
This is a short list of known issues:
- Humbug does initial test runs, logging and code coverage. Should allow user to do that optionally.
- Test classes (not tests) are run in a specific order, fastest first. Interdependent test classes may therefore fail regularly which will skew the results.
- Currently 100% PHPUnit specific, well 98.237%. There is an adapter where PHPUnit code is being shovelled.
- Certain test suite may make assumptions about having sole access to resources like /tmp which will cause errors when Humbug tries using same.
- Fine grained test ordering by speed (as opposed to large grained test class ordering) is awaiting implementation.
- Should test classes be used to carry non-PHPUnit dependent testing code (e.g. register_shutdown_function()), it may create issues when combined with one or more of Humbugs optimisations which assume a finished test really is finished.
Courtesy of Craig Davis who saw potential in a once empty repository :P.
.:::::::::::...
.::::::::::::::::::::.
.::::::::::::::::::::::::.
::::::::::::::::::::::::::::.
::::::::::::::::::::::::::::::: .,uuu ...
:::::::::::::::::::::::::::::::: dHHHHHLdHHHHb
....:::::::'` ::::::::::::::::::' uHHHHHHHHHHHHHF
.uHHHHHHHHH' ::::::::::::::`. uHHHHHHHHHHHHHP"
HHHHHHHHHHH `:::::::::::',dHHuHHHHHHHHP".g@@g
J"HHHHHHHHHP 4H ::::::::' u$$$.
".HHHHHHHHP" .,uHP :::::' uHHHHHHHHHHP"",e$$$$$c
HHHHHHHF' dHHHHf `````.HHHHHHHHHHP",d$$$$$$$P%C
.dHHHP"" JHHHHbuuuu,JHHHHHHHHP",d$$$$$$$$$e=,z$$$$$$$$ee..
"" .HHHHHHHHHHHHHHHHHP",gdP" ..3$$$Jd$$$$$$$$$$$$$$e.
dHHHHHHHHHHHHHHP".edP " .zd$$$$$$$$$$$"3$$$$$$$$c
`???""??HHHHP",e$$F" .d$,?$$$$$$$$$$$$$F d$$$$$$$$F"
?be.eze$$$$$".d$$$$ $$$E$$$$P".,ede`?$$$$$$$$
4."?$$$$$$$ z$$$$$$ $$$$r.,.e ?$$$$ $$$$$$$$$
'$c "$$$$ .d$$$$$$$ 3$$$.$$$$ 4$$$ d$$$$P"`,,
"""- "$$".`$$" " $$f,d$$P".$$P zeee.zd$$$$$.
ze. .C$C"=^" ..$$$$$$P".$$$'e$$$$$P?$$$$$$
.e$$$$$$$"="$f",c,3eee$$$$$$$$P $$$P'd$$$$"..::.."?$%
4d$$$P d$$$dF.d$$$$$$$$$$$$$$$$f $$$ d$$$" :::::::::.
$$$$$$ d$$$$$ $$$$$$$$$$$$$$$$$$ J$$",$$$'.::::::::::::
"$$$$$$ ?$$$$ d$$$$$$$$$$$$$$$P".dP'e$$$$':::::::::::::::
4$$$$$$c $$$$b`$$$$$$$$$$$P"",e$$",$$$$$' ::::::::::::::::
' ?"?$$$b."$$$$.?$$$$$$P".e$$$$F,d$$$$$F ::::::::::::::::::
"?$$bc."$b.$$$$F z$$P?$$",$$$$$$$ ::::::::::::::::::::
`"$$c"?$$$".$$$)e$$F,$$$$$$$' ::::::::::::::::::::
':. "$b...d$$P4$$$",$$$$$$$" :::::::::::::::::::::
':::: "$$$$$".,"".d$$$$$$$F ::::::::::::::::::::::
:::: be."".d$$$4$$$$$$$$F :::::::::::::::::::::::
:::: "??$$$$$$$$$$?$P" :::::::::::::::::::::::::
:::::: ?$$$$$$$$f .::::::::::::::::::::::::::::
:::::::`"????"".::::::::::::::::::::::::::::::