Обсуждение: [DOCS] to_char(): 'FM' also suppresses *trailing* zeroes

Erwin Brandstetter <brsaweda@gmail.com> writes:
> In Table 9-27. "Template Pattern Modifiers for Numeric Formatting" it says:
> FM | prefix fill mode (suppress leading zeroes and padding blanks) | FM9999

> In fact, 'FM' also suppresses *trailing* zeroes after the comma. To fix,
> this might be changed to:
>    suppress insignificant zeroes and padding blanks

Not necessarily.  A bit of experimentation says that it also matters
whether you use "0" or "9" as the format character:

So it appears to me that the bit you point out is flat out backwards;
what FM actually suppresses is trailing zeroes not leading zeroes.

I'm tempted to propose that in table 9-26, we need to write

        9       digit position (can be dropped if insignificant)

        0       digit position (cannot be dropped, even if insignificant)

and then in 9-27 say

        FM      fill mode: suppress trailing zeroes and padding spaces

Also, in between those two tables, I see

      * 9 results in a value with the same number of digits as there are
        9s. If a digit is not available it outputs a space.

This seems outright wrong per the above examples, and in any case is not
very useful since it doesn't explain the difference from "0".  Perhaps
rewrite as

        * 0 specifies a digit position that will always be printed,
          even if it contains a leading/trailing zero.  9 also specifies
          a digit position, but if it is a leading zero then it will be
          replaced by a space, while if it is a trailing zero and fill mode
          is specified then it will be deleted.

I think the root of the confusion is that the FM does for date/time formatting what the manual says (table 9-25):

FM | prefix fill mode (suppress leading zeroes and padding blanks) | FMMonth

test=# SELECT to_char(timestamp '0910-09-03 01:00:03.04', template)
test-# FROM  (
test(#    VALUES
test(#      ('YYYY-MM-DD HH24:MI:SS.US')
test(#    , ('FMYYYY-MM-DD HH24:MI:SS.US')
test(#    , ('FMYYYY-MM-DD FMHH24:MI:SS.US')
test(#    , ('FMYYYY-FMMM-FMDD FMHH24:FMMI:FMSS.FMUS')
test(#    ) t(template);
          to_char
----------------------------
 0910-09-03 01:00:03.040000
 910-09-03 01:00:03.040000
 910-09-03 1:00:03.040000
 910-9-3 1:0:3.040000

Note how trailing zeros are not suppressed for microseconds. (It would seem to make more sense to suppress those, though.)

The same explanation is given for numeric formatting (table 9-27):

FM prefix | fill mode (suppress leading zeroes and padding blanks) | FM9999

But it does not apply there - like you pointed out. The exact behavior seems to be:

For the pattern character 0 leading and trailing zeroes are always printed, no matter what.
For the pattern character 9 ...
    without FM modifier
        leading zeros are replaced with padding blanks (before the sign if it's there).
        trailing zeros after the decimal point are printed.

    with FM modifier
        leading and trailing zeros are removed (unless overruled by a 0).

There is additional blank padding for the sign if not printed - also removed with FM.