Обсуждение: "name" vs "alias" in datatype table

The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/16/datatype.html
Description:

I think there could be some clarification of what is a "name" vs. "alias" on
the datatypes table.  Right now, what's in the "Aliases" column is sometimes
postgres's internal type (e.g. `pg_catalog.int4`), and sometimes the "pretty
name", (e.g. `integer`).

How about if this table was laid out still as one row per type, but the
columns were "Names" (plural) and "Internal Name" (singular).  Names would
contain both "int" and "integer" (neither of which appear in `pg_type`) and
"Internal Name" would contain "pg_catalog.int4", the "real" name.  Then you
could also include arrays here, so that `text[]` would appear in the names
column and `pg_catalog._text` would appear in the "Internal Name" column.

On 23.11.23 21:51, PG Doc comments form wrote:
> I think there could be some clarification of what is a "name" vs. "alias" on
> the datatypes table.  Right now, what's in the "Aliases" column is sometimes
> postgres's internal type (e.g. `pg_catalog.int4`), and sometimes the "pretty
> name", (e.g. `integer`).

This appears not to be true.  `integer` is not listed as an alias.

Eric Hanson <eric@aquameta.com> writes:
> The larger point being, the "name" vs "alias" paradigm presented in this
> table does not accurately represent PostgreSQL, and conveys an inaccurate
> picture of the relationship between type names.  int4 is not an "alias".

I agree that this could be improved, mainly because it's far from
clear what the internal name of each type is (and there's at least
one case where the internal name is not shown at all).

I could see splitting this into three columns:

1. Preferred name (the standard's name, if it's a standard type)

2. Internal name (pg_type.typname), perhaps only if different from #1

3. Other aliases

However, the table is already pretty wide and so adding another
column might create formatting issues.

AFAICS the only candidates for "other aliases" are char, int,
and decimal.  Maybe we could handle those another way than reserving
a table column for them?  We could give them their own table rows,
or relegate them to footnotes.

The "serial" types need a bit more reflection too, since they
aren't truly types at all: there is no matching pg_type entry.
I'm not sure they belong here.

            regards, tom lane

On 2023-Nov-29, Tom Lane wrote:

> Eric Hanson <eric@aquameta.com> writes:
> > The larger point being, the "name" vs "alias" paradigm presented in this
> > table does not accurately represent PostgreSQL, and conveys an inaccurate
> > picture of the relationship between type names.  int4 is not an "alias".
> 
> I agree that this could be improved, mainly because it's far from
> clear what the internal name of each type is (and there's at least
> one case where the internal name is not shown at all).

Maybe we could split this into multiple *tables*.  The main one would be
what we already have except the aliases column is removed, and serial
types removed; the second table would list serial pseudo-types, without
aliases; the third one would have unofficial names (internal names and
other aliases) for types listed in the other two.

-- 
Álvaro Herrera               48°01'N 7°57'E  —  https://www.EnterpriseDB.com/
"The ability of users to misuse tools is, of course, legendary" (David Steele)
https://postgr.es/m/11b38a96-6ded-4668-b772-40f992132797@pgmasters.net