pandas to_datetime errors='ignore' deprecated (tested)

FutureWarning: errors='ignore' is deprecated and will raise in a future version: on pandas 3.1.0 it now raises ValueError: errors must be one of 'raise', or 'coerce'. We ran the same code on pandas 2.3.3 and 3.1.0 and tested the replacements, including pd.to_numeric.

1. errors="ignore" on pandas 2 and pandas 3

import warnings
import pandas as pd
warnings.simplefilter("always")

s = pd.Series(["2026-10-23", "not a date"])
try:
    print(pd.to_datetime(s, errors="ignore").tolist())
except ValueError as e:
    print("ValueError:", e)
try:
    pd.to_datetime(s)
except ValueError as e:
    print("ValueError:", str(e).splitlines()[0])
print(pd.to_datetime(s, errors="coerce").tolist())

Output with pandas 2.3.3:

/w/b1.py:7: FutureWarning: errors='ignore' is deprecated and will raise in a future version. Use to_datetime without passing `errors` and catch exceptions explicitly instead
  print(pd.to_datetime(s, errors="ignore").tolist())
['2026-10-23', 'not a date']
ValueError: time data "not a date" doesn't match format "%Y-%m-%d", at position 1. You might want to try:
[Timestamp('2026-10-23 00:00:00'), NaT]

Output with pandas 3.1.0:

ValueError: errors must be one of 'raise', or 'coerce'.
ValueError: time data "not a date" doesn't match format "%Y-%m-%d". You might want to try:
[Timestamp('2026-10-23 00:00:00'), NaT]

pandas 2.3.3 warns that errors="ignore" is deprecated; pandas 3 removed it and raises ValueError: errors must be one of 'raise', or 'coerce'. "Ignore" never meant "skip bad values": if any value failed, it returned the whole input unchanged, so the good date stayed a string too (first output line on pandas 2).

2. Replacements

import pandas as pd

s = pd.Series(["2026-10-23", "not a date"])
def to_dt_or_keep(x):                    # what errors="ignore" used to do
    try:
        return pd.to_datetime(x)
    except (ValueError, TypeError):
        return x
print(to_dt_or_keep(s).tolist())

parsed = pd.to_datetime(s, errors="coerce")   # or: keep per-value fallbacks
print(parsed.where(parsed.notna(), s).tolist())
print(s[parsed.isna()].tolist())              # see which values failed

Output (identical on pandas 2.3.3 and 3.1.0, no warnings):

['2026-10-23', 'not a date']
[Timestamp('2026-10-23 00:00:00'), 'not a date']
['not a date']

To keep the old all-or-nothing behaviour exactly, wrap the call in try/except as the warning text recommends. Usually what you want is errors="coerce": bad values become NaT and the good ones are parsed. s[parsed.isna()] lists the values that failed so you can fix the format (for example with format= or dayfirst=True). Mixing Timestamps and strings in one column, as in the second output line, gives an object column, so only do that for display.

3. Same change in pd.to_numeric

import warnings
import pandas as pd
warnings.simplefilter("always")

s = pd.Series(["1", "x"])
try:
    print(pd.to_numeric(s, errors="ignore").tolist())
except ValueError as e:
    print("ValueError:", e)
print(pd.to_numeric(s, errors="coerce").tolist())

Output with pandas 2.3.3:

/w/b3.py:7: FutureWarning: errors='ignore' is deprecated and will raise in a future version. Use to_numeric without passing `errors` and catch exceptions explicitly instead
  print(pd.to_numeric(s, errors="ignore").tolist())
['1', 'x']
[1.0, nan]

Output with pandas 3.1.0:

ValueError: invalid error value specified
[1.0, nan]

to_numeric(errors="ignore") went the same way, with a vaguer pandas 3 message: invalid error value specified. Use errors="coerce" or try/except here too.

Related: all pandas 3 breaking changes · FutureWarning: concat with empty or all-NA entries · SettingWithCopyWarning · Error tokenizing data.

Written with AI assistance. Measured on 2026-10-23 in the python:3-alpine Docker image (Python 3.14) with pandas 2.3.3 and pandas 3.1.0. Each block was run verbatim on both versions with warnings.simplefilter("always") and the output pasted unedited. PlainHeaders is our extension; Python and pandas are free.