On 2015/12/27 3:11, Tom Lane wrote:
> I wrote:
>> Amit Langote <langote_amit...@lab.ntt.co.jp> writes:
>>> Any specific reason why it doesn't spell out typmods in the above detail
>>> message?
>> * There's a rough policy in the parser to prefer TypeNameToString
>> when complaining about a TypeName input, rather than reconstructing
>> the type name from the OID.  The reason for this is that we'd rather
>> complain about the type name as entered, not the canonical type name
>> --- for example, if the user typed "float8" it might be a bit confusing
>> if the parser then complains about "double precision".
>> I'm not entirely sure though that that argument should be applied
>> to this particular case, because the other type referred to will
>> certainly get displayed in canonical form.
> On reflection, I think trying to spell both types according to the same
> rules will be the least confusing behavior here.
>> So we could either apply your patch as written, or we could replace
>> only the format_type_be calls with format_type_with_typemod, and
>> then fix TypeNameToString so that it will show the typmod if any.
>> (We'd need to consider whether that behavior is OK for all callers.)
>> Even if we decide this particular case is best handled by presenting
>> canonical type names on both sides, maybe it would be wise to look
>> into whether TypeNameToString should be changed for other callers.
> I looked through the other call sites for TypeNameToString and
> TypeNameListToString.  In none of them does it seem useful to include any
> typmod info in the printout, and in many of them it would be positively
> misleading (e.g., functions do not care about typmod decorations on their
> argument types).  So we should not change the behavior of those functions.
> Perhaps down the road there'll be a use for "TypeNameAndTypmodToString",
> but I don't feel a need for it today.
> So I am thinking your patch is good as proposed, ie, let's use
> format_type_with_typemod here.

I agree. Thanks for adding the tests.


Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:

Reply via email to