Thread: ...
Hi all, I have a problem with metadata.........just read this ....... Let us simply suppose a table "test" with 2 fileds "name (varchar(10))" and "age (numeric)" and there b values as name="abc",age="20". Now in a function i need to develop a list where the column header info has to b made in this format i.e., as "column name, column type, column width"......... am getting the column name using PQfname function, column type using PQftype and column size using PQfsize....... am also able to get each filed values n their size correctly.......... but the problem now is ......when i use PQfsize to get the column header size then i get a -1 if the field is a variable length in nature else i get the exact size........i.e., in the above declared table for both the name and age being variable fields am getting -1 for their size.......now i need some mechanism to get 10 for "name" header and number of bytes allocated for "age"(30,6).......is their any way to overcome this problem? Is this a draw back in Postgre? Is there any way i can get the exact size that i allocated when creating the table.....infact most of the other databases do provide APIs with not having this problem........can anyone help me, please......... Shra
Shra <shravan@yaskatech.com> writes:
> [ where to get declared-length info for var-length columns ]
This is encoded in the atttypmod field (see PQfmod). I'd recommend
using format_type() to decode the info, but if you don't mind possibly
having to change your code from time to time, you could just wire
knowledge of the interpretation of typmod into your code.
In future, kindly be more selective about the lists you cross-post to.
I generally make a practice of ignoring questions posted to the bugs
list, for example.
regards, tom lane
Hi tom, how to decode the atttypmod field........when i allocate 10 it shows it as 14 n for 20 it shows it as 24.......my problem is not just with varchar........it is for all fields that has variable length .........the same atttypmod returns -1 for integer? am confused........ I want the size allocated for each field i a table whether its a var-length or fixed length........ Shra On Saturday 27 April 2002 20:37, you wrote: > Shra <shravan@yaskatech.com> writes: > > [ where to get declared-length info for var-length columns ] > > This is encoded in the atttypmod field (see PQfmod). I'd recommend > using format_type() to decode the info, but if you don't mind possibly > having to change your code from time to time, you could just wire > knowledge of the interpretation of typmod into your code. > > In future, kindly be more selective about the lists you cross-post to. > I generally make a practice of ignoring questions posted to the bugs > list, for example. > > regards, tom lane -- Shra
Hi Hannu, > -1 is given for types that are of fixed size and whose length can be > read from pg_type.typlen for that type. I don't think so...jsut look into this file pq_type.h.... it says......... *****************************************************************typlen is the number of bytes we use to represent a valueof this type, e.g. 4 for an int4. But for a variable length type, typlen is -1 ***************************************************************** -1 is for variable length n not for fixed length.........this point is very clear even in documentation.......... now how to find length for a numeric, varchar or anyother one that has variable length where the system PQfsize returns -1.........? As tom said.....The type is encoded in the atttypmod field (see PQfmod) and recommended using format_type()..... but when this is used, it returns -1 for integer , real n other fixed datatypes ......... any better way , please......... Shra
On Mon, 2002-04-29 at 13:16, Shra wrote: > Hi Hannu, > > > -1 is given for types that are of fixed size and whose length can be > > read from pg_type.typlen for that type. > > I don't think so...jsut look into this file pq_type.h.... it says......... > ***************************************************************** > typlen is the number of bytes we use to represent a value of this type, e.g. > 4 for an int4. But for a variable length type, typlen is -1 > ***************************************************************** > > -1 is for variable length n not for fixed length.........this point is very > clear even in documentation.......... Yes, it in pg_type.typlen it is -1 is for variable length and actual length for fixed-length types in pg_attribute.attypmod it is -1 for _fixed_ length types and actual length for variable length types (actual length = defined length + 4 bytes of length bytes) > now how to find length for a numeric, varchar or anyother one that has > variable length where the system PQfsize returns -1.........? > > As tom said.....The type is encoded in the atttypmod field (see PQfmod) and > recommended using format_type()..... > but when this is used, it returns -1 for integer , real n other fixed > datatypes ......... so do as Tom said - if PQfsize returns -1 then use PQfmod else use PQfsize ----------- Hannu
Shra <shravan@yaskatech.com> writes:
> As tom said.....The type is encoded in the atttypmod field (see PQfmod) and
> recommended�using format_type().....
> but when this is used, it returns -1 for integer , real n other fixed
> datatypes .........
The interpretation of typmod is datatype-specific. That's why I
recommended using format_type. But if you want to know what format_type
knows, look at the source code (src/backend/utils/adt/format_type.c).
regards, tom lane