测试数据类型时，在for循环中使用正则表达式的最佳方法是什么？

Question

What is the best way to use regex in a for loop while testing for data type?

For context, I'm looping over large unclean data sets with multiple data types and need to find extensions of strings, if they exist. Small changes to my code, like converting values to string costs me minutes.

I read through this question Python: How to use RegEx in an if statement? but couldn't find a way of testing for a match without first converting to a string.

Values:

vals = [444444, '555555-Z01']
pattern = re.compile('[-]*[A-Z]{1}[0-9]{2}$')
# new_vals = [444444, 555555]

Slow method: (2.4 µs ± 93.6 ns per loop)

new_vals = []
for v in vals:
    if type(v)==str:
        if pattern.search(v) is not None:
            new_v = pattern.findall(v)[0].replace('-','')
            new_vals.append(new_v)
    else:
        new_vals.append(v)

Fast method: (1.84 µs ± 34.7 ns per loop)

f = lambda x: x if type(x)!=str else pattern.findall(x)[0].replace('-','')

new_vals = []
for v in vals:
    new_vals.append(f(v))

Unsucessful Method:

new_vals = []
for v in vals:
    if ((type(v)==str) & (pattern.search(v) is not None)):
        new_vals.append(v)

Error:

TypeError: expected string or bytes-like object

Answer 1

I tried to beat your attempts using try/except blocks but the exception handling seems to take too much time. So much for "better ask forgiveness than permission" ...

Your last attempt is the most promising, if you just change & by and , because & is the logical and and doesn't short circuit.

I'll go for this, in a list comprehension to speed it up slightly more, and drop the is not None test which is useless since if search succeeds, it returns a regex object, which is truthy:

new_vals = [v for v in vals if type(v)==str and pattern.search(v)]

or with isinstance (same speed, tests subclasses of str too):

new_vals = [v for v in vals if isinstance(v,str) and pattern.search(v)]