caikit.interfaces.ts.data_model.backends.util

Internal utilities for supporting backend implementations

Functions

timezoneoffset(→ int)	Returns the timezone offset (in seconds)
pd_timestamp_to_seconds(→ float)	Extract the seconds-since-epoch representation of the timestamp
strip_periodic(→ pandas.DataFrame)	Removes the first instance of a periodic timestamp info

Module Contents

caikit.interfaces.ts.data_model.backends.util.timezoneoffset(adatetime: datetime.datetime) → int

Returns the timezone offset (in seconds) for a given datetime object relative to the local system’s time.

Args:

adatetime (datetime): a date of interest.

Returns:

int: offset in seconds (can be negative)

caikit.interfaces.ts.data_model.backends.util.pd_timestamp_to_seconds(ts) → float

Extract the seconds-since-epoch representation of the timestamp

NOTE: The pandas Timestamp.timestamp() function returns a different value

than Timestamp.to_pydatetime().timestamp()! Since we want this to round-trip with python datetime, we want the latter. They both claim to be POSIX, so something is missing leap-something!

caikit.interfaces.ts.data_model.backends.util.strip_periodic(input_df: pandas.DataFrame, ts_col_name: str | None = None, create_copy=True) → pandas.DataFrame

Removes the first instance of a periodic timestamp info (because spark doesn’t like these when constructing a pyspark.sql.DataFrame.) If no periodic timestamp values can be found, input_df is returned as is. This method is always a no-op if input_df is not a native pandas.DataFrame.