[Python] Replace values in series with first regex group if found, if not leave nan

Say I have a dataframe such as

df = pd.DataFrame(data = {'col1': [1,np.nan, '>3', 'NA'], 'col2':["3.","<5.0",np.nan, 'NA']})


out:
col1 col2
0 1 3.
1 NaN <5.0
2 >3 NaN
3 NA NA


What I would like is to strip stuff like "<" or ">", and get to floats only

out:
col1 col2
0 1 3.
1 NaN 5.0
2 3 NaN
3 NaN NaN


I thought about something like

df['col2'].replace({'.*' : r"[-+]?(?:\d*\.*\d+)"}, regex=True, inplace=True)


the idea being, replace anything with the regex for float, (I think), but this fails

error: bad escape \d at position 8


I tried along the lines of

df['col2'].replace({r"[-+]?(?:\d*\.*\d+)": r"\1"}, regex=True, inplace=True)


assuming ">"or "<" come before the float number, but then this fails (does not catch anything) if the field is a string or np.nan.

Any suggestions please?

Continue reading...