FORMAT with exceeding number of digits in second's fraction

[aafemt]

If CAST FORMAT into TIMESTAMP had fully implemented FF5+ formats, what result would you expect from CAST('00:00:00.000099999' AS TIME FORMAT 'HH:MI:SS.FF9') given that Firebird TIME can keep only four digits as the fraction?

Rounding or truncation?

[mrotteveel]

Personally, I would rather see that FF5 - FF9 are removed.

[aafemt]

And that why "if" in the starting post.

[TreeHunter9]

It is documented.

Maybe you have confused 1.DATETIME TO STRING with 2.STRING TO DATETIME in README.cast.format.md?

[aafemt]

It is documented that they are not implemented.

This discussion is about the expected behavior in the case if they had been implemented.

[TreeHunter9]

This discussion is about the expected behavior in the case if they had been implemented.

I know, I was just responding to Mark about removing FF5 - FF9.

Also, I currently fixing incorrect behavior of FF1 - FF4, and I thought about "fractional precision" part in description of these patterns. Oracle implements FF1 - FF9 in a way that 1 - 9 is not kind of a "fractional precision", it's more like "fractional limitation", for example: to_timestamp('30.09', 'SS.FF1') will not give us 30.1, it will throw an error. So the question is, should we also throw an error too, or round the result to the given "fractional precision" pattern?

[aafemt]

It is an obvious case of string not matching pattern, so throwing an error is the only logical behavior for the case when the string has more digits than the pattern expects.