DidYouMean-Python (aka BetterErrorMessages)

Logic to have various kind of suggestions in case of errors (NameError, AttributeError, ImportError, TypeError, ValueError, SyntaxError, MemoryError, OverflowError, IOError, OSError).

Inspired by "Did you mean" for Ruby (Explanation,Github Page), this is a simple implementation for/in Python. I wanted to see if I could mess around and create something similar in Python and it seems to be possible.

This project is not maintained anymore

As of October 2021, this project is not maintained anymore. Indeed, improvements to the exception messages have been implemented in Python 3.8, 3.9, 3.10 and 3.11 making this project useless (which is a good thing).

If you are interested in projects aiming to make exception messages better, please refer to thelist of projects below.In particularfriendly-traceback/friendly-traceback.

I'll try to ensure that the test suite runs fine on all Python versions as it was useful in the past to catch issues in the Python project because things got released.

Usage

Once the package is installed (see below), the logic adding suggestions can be invoked in different ways:

• hook onsys.excepthook:just calldidyoumean_enablehook()and you'll have the suggestions for any uncaught exception:

>>> abc = 3
>>> abcd
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'abcd' is not defined
>>> didyoumean.didyoumean_api.didyoumean_enablehook()
>>> abcd
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'abcd' is not defined. Did you mean 'abc' (local)?

• post-mortem functiondidyoumean_postmortem()on the last uncaught exception during interactive sessions:

>>> abc = 3
>>> abcd
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'abcd' is not defined
>>> didyoumean.didyoumean_api.didyoumean_postmortem()
NameError( "name 'abcd' is not defined. Did you mean 'abc' (local)?",)

• context managerdidyoumean_contextmanager():

>>> with didyoumean.didyoumean_api.didyoumean_contextmanager():
... abcd
...
Traceback (most recent call last):
File "<stdin>", line 2, in <module>
NameError: name 'abcd' is not defined. Did you mean 'abc' (local)

• decorator: just add the@didyoumeandecorator before any function (themain()could be a good choice) and you'll have the suggestions for any exception happening through a call to that method.

>>> @didyoumean.didyoumean_api.didyoumean_decorator
... def foo(): return abcd
...
>>> foo()
Traceback (most recent call last):
File "<stdin>", line 2, in foo
NameError: global name 'abcd' is not defined. Did you mean 'abc' (global)?

The API does not look great and may be updated in the near future.

Example

More examples can be found from the test filedidyoumean/didyoumean_sugg_tests.py.

NameError

Fuzzy matches on existing names (local, builtin, keywords, modules, etc)

defmy_func(foo,bar):
returnfoob
my_func(1,2)
#>>> Before: NameError( "global name 'foob' is not defined",)
#>>> After: NameError( "global name 'foob' is not defined. Did you mean 'foo' (local)?",)

leng([0])
#>>> Before: NameError( "name 'leng' is not defined",)
#>>> After: NameError( "name 'leng' is not defined. Did you mean 'len' (builtin)?",)

importmath
maths.pi
#>>> Before: NameError( "name 'maths' is not defined",)
#>>> After: NameError( "name 'maths' is not defined. Did you mean 'math' (local)?",)

passs
#>>> Before: NameError( "name 'passs' is not defined",)
#>>> After: NameError( "name 'passs' is not defined. Did you mean 'pass' (keyword)?",)

defmy_func():
foo=1
foob+=1
my_func()
#>>> Before: UnboundLocalError( "local variable 'foob' referenced before assignment",)
#>>> After: UnboundLocalError( "local variable 'foob' referenced before assignment. Did you mean 'foo' (local)?",)

Checking if name is the attribute of a defined object

classDuck():
def__init__(self):
quack()
defquack(self):
pass
d=Duck()
#>>> Before: NameError( "global name 'quack' is not defined",)
#>>> After: NameError( "global name 'quack' is not defined. Did you mean 'self.quack'?",)

importmath
pi
#>>> Before: NameError( "name 'pi' is not defined",)
#>>> After: NameError( "name 'pi' is not defined. Did you mean 'math.pi'?",)

Looking for missing imports

string.ascii_lowercase
#>>> Before: NameError( "name 'string' is not defined",)
#>>> After: NameError( "name 'string' is not defined. Did you mean to import string first?",)

Looking in missing imports

choice
#>>> Before: NameError( "name 'choice' is not defined",)
#>>> After: NameError( "name 'choice' is not defined. Did you mean 'choice' from random (not imported)?",)

Special cases

assertj**2==-1
#>>> Before: NameError( "name 'j' is not defined",)
#>>> After: NameError( "name 'j' is not defined. Did you mean '1j' (imaginary unit)?",)

AttributeError

Fuzzy matches on existing attributes

lst=[1,2,3]
lst.appendh(4)
#>>> Before: AttributeError( "'list' object has no attribute 'appendh'",)
#>>> After: AttributeError( "'list' object has no attribute 'appendh'. Did you mean 'append'?",)

importmath
math.pie
#>>> Before: AttributeError( "'module' object has no attribute 'pie'",)
#>>> After: AttributeError( "'module' object has no attribute 'pie'. Did you mean 'pi'?",)

Trying to find method with similar meaning (hardcoded)

lst=[1,2,3]
lst.add(4)
#>>> Before: AttributeError( "'list' object has no attribute 'add'",)
#>>> After: AttributeError( "'list' object has no attribute 'add'. Did you mean 'append'?",)

lst=[1,2,3]
lst.get(5,None)
#>>> Before: AttributeError( "'list' object has no attribute 'get'",)
#>>> After: AttributeError( "'list' object has no attribute 'get'. Did you mean 'obj[key]' with a len() check or try: except: KeyError or IndexError?",)

Detection of mis-used builtins

lst=[1,2,3]
lst.max()
#>>> Before: AttributeError( "'list' object has no attribute 'max'" )
#>>> After: AttributeError( "'list' object has no attribute 'max'. Did you mean 'max(list)'?" )

Period used instead of comma

a,b=1,2
max(a.b)
#>>> Before: AttributeError( "'int' object has no attribute 'b'" )
#>>> After: AttributeError( "'int' object has no attribute 'b'. Did you mean to use a comma instead of a period?" )

ImportError

Fuzzy matches on existing modules

frommathsimportpi
#>>> Before: ImportError('No module named maths',)
#>>> After: ImportError( "No module named maths. Did you mean 'math'?",)

Fuzzy matches on elements of the module

frommathimportpie
#>>> Before: ImportError('cannot import name pie',)
#>>> After: ImportError( "cannot import name pie. Did you mean 'pi'?",)

Looking for import from wrong module

fromitertoolsimportpi
#>>> Before: ImportError('cannot import name pi',)
#>>> After: ImportError( "cannot import name pi. Did you mean 'from math import pi'?",)

TypeError

Fuzzy matches on keyword arguments

defmy_func(abcde):
pass
my_func(abcdf=1)
#>>> Before: TypeError( "my_func() got an unexpected keyword argument 'abcdf'",)
#>>> After: TypeError( "my_func() got an unexpected keyword argument 'abcdf'. Did you mean 'abcde'?",)

Confusion between brackets and parenthesis

lst=[1,2,3]
lst(0)
#>>> Before: TypeError( "'list' object is not callable",)
#>>> After: TypeError( "'list' object is not callable. Did you mean 'list[value]'?",)

defmy_func(a):
pass
my_func[1]
#>>> Before: TypeError( "'function' object has no attribute '__getitem__'",)
#>>> After: TypeError( "'function' object has no attribute '__getitem__'. Did you mean 'function(value)'?",)

ValueError

Special cases

'Foo{}'.format('bar')
#>>> Before: ValueError('zero length field name in format',)
#>>> After: ValueError('zero length field name in format. Did you mean {0}?',)

importdatetime
datetime.datetime.strptime("%d %b %y","30 Nov 00")
#> Before: ValueError( "time data '%d %b %y' does not match format '30 Nov 00'",)
#> After: ValueError( "time data '%d %b %y' does not match format '30 Nov 00'. Did you mean to swap value and format parameters?",)

SyntaxError

Fuzzy matches when importing fromfuture

from__future__importdivisio
#>>> Before: SyntaxError('future feature divisio is not defined',)
#>>> After: SyntaxError( "future feature divisio is not defined. Did you mean 'division'?",)

Various

return
#>>> Before: SyntaxError( "'return' outside function", ('<string>', 1, 0, None))
#>>> After: SyntaxError( "'return' outside function. Did you mean to indent it, 'sys.exit([arg])'?", ('<string>', 1, 0, None))

MemoryError

Search for a memory-efficient equivalent

range(99999999999)
#>>> Before: MemoryError()
#>>> After: MemoryError( ". Did you mean 'xrange'?",)

OverflowError

Search for a memory-efficient equivalent

range(999999999999999)
#>>> Before: OverflowError('range() result has too many items',)
#>>> After: OverflowError( "range() result has too many items. Did you mean 'xrange'?",)

OSError/IOError

Suggestion for tilde/variable expansions

os.listdir('~')
#>>> Before: OSError(2, 'No such file or directory')
#>>> After: OSError(2, "No such file or directory. Did you mean '/home/user' (calling os.path.expanduser)?" )

RuntimeError

Suggestion to avoid reaching maximum recursion depth

globalrec
defrec(n):returnrec(n-1)
rec(0)
#>>> Before: RuntimeError('maximum recursion depth exceeded',)
#>>> After: RuntimeError('maximum recursion depth exceeded. Did you mean to avoid recursion (cf http://neo Python ic.blogspot.fr/2009/04/tail-recursion-elimination.html), increase the limit with `sys.setrecursionlimit(limit)` (current value is 1000)?',)

Installation

The package is available onPypiasBetterErrorMessages.

Installation can be done from the package index withpip install BetterErrorMessages.

Installation from sources can be done just as easily:

git clone https://github /SylvainDe/DidYouMean-Python.git
cd DidYouMean-Python
git install.

Making things automatic in your interactive sessions

You can have the suggestions automatically in your interactive sessions by adding the following code in your${PYTHONSTARTUP} file:

try:
import didyoumean
except ImportError:
print( "Did you mean to install BetterErrorMessages first (`pip install BetterErrorMessages`)" )
else:
didyoumean.didyoumean_api.didyoumean_enablehook()

Implementation

All external APIs (decorator, hook, etc) use the same logic behind the scene. It works in a pretty simple way: when an exception happens, we try to get the relevant information out of the error message and of the backtrace to find the most relevant suggestions. To filter the best suggestions out of everything in case of fuzzy match, I am currently usingdifflib.

See also (similar projects/ideas)

Projects:

• "Did you mean" for Ruby (yuki24/did_you_mean):source for the idea.

• dutc/didyoumean:a quite similar project developed in pretty much the same time. A few differences though: written in C, works only for AttributeError, etc.

• nvbn/TheF*ck:Correct and execute your previous shell command.

• asweigart/PyDidYouMean:Improve "file/command not found" errors with suggestions.

• Qix-/better-exceptions:Pretty and useful exceptions in Python

• danrobinson/tracestack:Search your Python error messages on the web.

• cfbolz/syntaxerrors:Python parser that can recover from errors (also, cfbolz added many suggestions for errors to PyPy).

• friendly-traceback/friendly-traceback:Replace standard traceback by something easier to understand.

• Did You Mean in Perl

• Commit in iPythonto add suggestions in case of errors

Discussions:

• PEP 473: Adding structured data to built-in exceptions.

• Ideas from the Python Ideas mailing list:"Improve error message when missing 'self' in method definition","Better error messages"part 1andpart 2

• InRaymond Hettinger's PyconCA keynote,the part about thehintbuiltin (at 14 minutes) looks a lot likedidyoumean_postmortem.

Contributing

Feedback is welcome, feel free to:

• send me an email for any question/advice/comment/criticism
• open issues if something goes wrong (please provide at least the version of Python you are using).

Also, pull-requests are welcome to:

• fix issues
• enhance the documentation
• improve the code
• bring awesomeness

As for the technical details:

• this is under MIT License: you can do anything you want as long as you provide attribution back to this project.
• I try to followPEP 8andPEP 257as much as possible. Compliancy is checked during continuous integration using thepycodestyleandpep257checkers.
• I try to have most of the code covered by unit tests.
• I try to write the code in such a way that it works on all Python versions from 2.6 (included).