Bruce Momjian wrote:
> I talked to Andrew on IRC and we went over the open CSV issues.
> We talked about how we could do quoting for zipcode in TEXT fields and
> not quote true numeric values without hardcoding datatypes into the
> system.  The most tricky case was NUMERIC vs. TEXT with zipcodes. 
> NUMERIC and TEXT have almost identical pg_type entries, so there is
> nothing there to help us.
> I found parse_coerce.c::TypeCategory(), which contains information about
> which data type oids are in which grouping, e.g. DATETIME, STRING,
> NUMERIC, etc.  It seems that function, if called with
> pg_type.typbasetype could help determine if quotes should be used.  My
> idea is to skip quotes for NUMERIC and DATETIME types, and quote
> everything else.  This means that user-defined types will always be
> output with quotes, which is probably OK.
> So, for open CSV items we have:
>       o  add oid dump/reload
>       o  handle loading of comma-comma into NOT NULL collumns
>       o  handle quoting of TEXT type with zipcodes, etc.

Oh, one more item.  CSV needs a way to output date values in Datestyle
format, for use by spreadsheets that understand mm/dd/yy dates, but not
ISO dates.  This will not be on by default.

  Bruce Momjian                        |
  [EMAIL PROTECTED]               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
      joining column's datatypes do not match

Reply via email to