Native transformations between numerals and int

[mkgrgis]

Feature requests

1. Now there is only select roman_in(to_char(generate_series(0, 10001), '99999')::cstring) transformation. What about roman_in(smallint) or roman_in(int)?

2. If there is unicode-encoding (see some constants here) what about ↁ = 5000, ↂ = 10000, ↇ = 50000, ↈ = 100000 for input and output? Also there is less compact forms IƆƆ = 5000, CCIƆƆ = 10000, IƆƆƆ = 50000, CCCIƆƆƆ = 100000 at least for input. Note: unfortunately SELECT 'ↈ'::roman is incorrect now.

3. In Unicode also there is special roman forms

   U+2160 | Ⅰ 2160 | Ⅴ 2164 | Ⅹ 2169 | Ⅼ 216C | Ⅽ 216D | Ⅾ 216E | Ⅿ 216F
   U+2170 | ⅰ 2170 | ⅴ 2174 | ⅹ 2179 | ⅼ 217C | ⅽ 217D | ⅾ 217E | ⅿ 217F
   U+2180 | ↀ 2180 | ↁ 2181 | ↂ 2182 | Ↄ 2183 | ↅ 2185 | ↆ 2186 | ↇ 2187 | ↈ 2188

   But it's incorrect input now: SELECT 'Ⅹ'::roman. What about support of this forms at least for input?

4. Cast functions to smallint and int. There is internal numeric presentation, but no output.

Note: I comes with postgresql-numeral package from PGDG apt sever. Thanks for usefully SQL utility!

[df7cb]

Hi,

1. the idea is that we override as few PG functions as possible, work with bigint, and rely on casts from the other numeric types. When implicit casts don't work, use explicit casts:

create table r (n smallint);
insert into r values (1);
select n::roman from r;
FEHLER:  42846: kann Typ smallint nicht in Typ roman umwandeln
select n::bigint::roman from r;
 n 
───
 I

2. interesting, I didn't know these symbols yet. Will consider that the next time I touch the extension.

3. use type casts.

[mkgrgis]

Thanks, @df7cb !

1. select generate_series(0, 10001)::bigint::roman works fine, but more SQL:2016 compatible is direct functions for int and smallint. Will it hard for You to add this functions by bigint example? Maybe select generate_series(0, 10001)::bigint::roman will be interesting as example in documentation / README ?
2. Lies Sie bitte Große Zahlen und
3. Darstellung in Unicode
4. select r::bigint from t works, but also more SQL:2016 compatible are direct casts to int and smallint. Will it hard for You to add this functions by bigint example?

My notes about int and smallint in context SQL:2016 are also applicable to zahl and numeral.