scikit-learn random_state check

rcheck.py

# rcheck.py: check sklearn test files for class/function invocations that
# lack the optional `random_state` argument. For each file a missing invocation
# is found, prints out the class/function name followed by line number(s)
#
# Usage (from scikit-learn base directory) 
# $ find * -type f | grep "/test_.*py$" | python rcheck.py
import fileinput
import inspect
import importlib
import re

ARGNAME = 'random_state'

def get_estimators(mod_name):
    # Get name of imported callables that takes a `random_state` arg
    module = importlib.import_module(mod_name)
    random_estimators = []
    for el in dir(module):
        obj = getattr(module, el)
        try:
            if ARGNAME in inspect.signature(obj).parameters:
                random_estimators.append(el)
        except:
            pass

    return random_estimators

def process_file(f_name, random_estimators):
    # For each of the estimators in `random_estimators` search
    # the source file `f_name` for invocations and see if these 
    # lack a `random_state` argument. This is done in a simple regex
    # that matches up to the first closing parenthesis, so will miss
    # invocations that contain function calls (i.e. nested parenthes)
    # or invoke the estimator in other ways (through a class factory
    # or maybe with **kwrds)
    nmissing = 0
    msg = "\t{0} {1}"
    with open(f_name, 'r') as f:
        text = f.read()
        for estimator in random_estimators:
            linenums = []
            for m in re.finditer(estimator + '\([^\)]*', text):
                match = text[m.start():m.end()]
                if match.find(ARGNAME) == -1:
                    lineno = text.count('\n', 0, m.start()) + 1
                    linenums.append(lineno)
                    nmissing += 1
            if nmissing != 0 and estimator == random_estimators[0]:
                print(f_name)
            if linenums: 
                print(msg.format(estimator, linenums))
    return nmissing

nmissing = 0
nfiles = 0
for f in fileinput.input():
    f_name = f.rstrip()
    mod_name = f_name.replace('/', '.').replace('.py', '')
    random_estimators = get_estimators(mod_name)
    nmissing += process_file(f_name, random_estimators)
    nfiles += 1

print ("")
print ("Found {0} missing in {1} files".format(nmissing, nfiles))