[Firebird-devel] Type of nod_udf

[Claudio V. C. <cv...@us...>]

Neil, regarding your latest changes through fb-checkins:

string = (STR) node->nod_arg [e_udf_name];
+ /* how are we supposed to tell which type of nod_udf this is ?? */
+ if (string->str_data[0] == NULL)
+ PRINTF("%s\"\n",((UDF) node->nod_arg[0])->udf_name);
+ else
+ PRINTF ("%s\"\n", string->str_data);

What's the pattern for the two types of UDFs? I thought the name always
comes in a nod_arg at index e_udf_name, but your code proves other fact.
Isn't there another place in the code when the udf name is pushed into the
BLR stream? Since e_udf_name is zero in practice, I understand your problem
is between
char *s = ((STR) node->nod_arg [0])->str_data;
and
char *s = ((UDF)node->nod_arg[0])->udf_name

Strange problem. One possible source could be our dsql parser that might be
generating different nodes. The other explanation is that you're in a map
when you have aggregate functions or sub-selects, have a look at pass1.c
when a name is assigned to each column, there's a long if-else-if
construction.

In practice, you can determine the type of the node by looking at
nod->nod_type (high level type)
and
nod->nod_header.blk_type (what the allocator knows)

The types that the DSQL allocator knows are in dsql\blk.h, you could test
such info; there are type_udf and type_str. I'm still wondering why other
places of the code don't seem to have that same problem.

C.
---------
Claudio Valderrama C.
Ingeniero en Informática - Consultor independiente
http://www.cvalde.com - http://www.firebirdSQL.org