ENH: Raise TypeError for offsets which are not supported as period frequency

[natmokval]

• closes BUG: QuarterBegin Does not work with Period #55785
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.

Added to constructor of Period check if frequency is supported as period frequency, if not a ValueError is raised.

The reason: so far we get
AttributeError: 'pandas._libs.tslibs.offsets.QuarterBegin' object has no attribute '_period_dtype_code'

[MarcoGorelli]

instead of hard-coding the offsets and checking this, could we just do

if not hasattr(freq, '_period_dtype_code'):
    raise ValueError(...)

?

[natmokval]

Thank you, I corrected the condition and the error message. We have _period_dtype_code in class CustomBusinessDay (see link) that is why I excluded CustomBusinessDay.

I thought, isn't better to use TypeError instead of ValueError?
Should we add a note to v2.2.0? Something like Bug in incorrectly allowing construction of :class:'Period' ...

[MarcoGorelli]

thanks - yeah TypeError is probably better

and yes, a 2.2.0 whatsnew note would be good

maybe we can just do

try:
    period_dtype_code = freq._period_dtype_code
except (AttributeError, TypeError):
    # AttributeError: _period_dtype_code might not exist
    # TypeError: _period_dtype_code might intentionally raise
    raise TypeError(f"{(type(freq).__name__)} is not supported as period frequency"

and then reuse period_dtype_code further down where there's freq._period_dtype_code?

[natmokval]

thanks, I did as you suggested. It works very well.

[MarcoGorelli]

Nice work! minor comment on the whatsnew note but the rest looks good on green

[MarcoGorelli]

I think we're giving a better error message here, rather than fixing a bug - could we rephrase to reflect that, and put this in "other enhancements"?

something like

- Improved error message when constructing :class:`Period` with invalid offsets such as "QS" (:issue:`55785`)

[natmokval]

agree, I did as you suggested.

[mroeschke]

Thanks @natmokval

[cancan101]

@natmokval, does this PR also close: #5091 ?

[natmokval]

@natmokval, does this PR also close: #5091 ?

@cancan101, I think, this PR doesn't close #5091. We still have the same behavior. For example, frequency changes from 'ME' to 'MS':

>>> ts = pd.date_range('1/1/2012', periods=4, freq=pd.offsets.MonthEnd())
>>> per = ts.to_period()
>>> ts_per = per.to_timestamp()
>>> ts
DatetimeIndex(['2012-01-31', '2012-02-29', '2012-03-31', '2012-04-30'], dtype='datetime64[ns]', freq='ME')
>>> per
PeriodIndex(['2012-01', '2012-02', '2012-03', '2012-04'], dtype='period[M]')
>>> ts_per
DatetimeIndex(['2012-01-01', '2012-02-01', '2012-03-01', '2012-04-01'], dtype='datetime64[ns]', freq='MS')