[postgis-users] Is that possible a function to behave differently inside and outside another main function code?

[Kevin Neufeld]

Are you sure you want to use "UNION" and not "UNION ALL"?  The former 
will remove duplicates, the latter does not.  It's conceivable that when 
UNIONed, the three SELECT st_X clauses will return a single value.  
Collected and put through ST_LineFromMultiPoint would probably result in 
a single point line (depending on which version of PostGIS you are using 
- the newer versions will ERROR with "geometry requires more points").

Hope that helps,
Kevin

rodrigosperb wrote:
> Hello,
>
> I have a bit of a problem that is sort of driving me crazy. I need to
> perform an "addition of two (mathematical) functions". I represent them as
> linestrings in my solution, and it is part of another bigger function. The
> code is as follows:
>
> ...header...
> SELECT st_LineFromMultiPoint(st_Collect(st_MakePoint(xy.t,xy.at)))
> FROM (SELECT q.t, dr_delay_value($1,q.t) +
> dr_delay_value($3,dr_delay_value($1,q.t)) AS at
>       FROM (SELECT st_X(st_PointN($1,n)) AS t
> 	    FROM generate_series(1,st_NumPoints($1)) AS h(n)
> 	    UNION
> 	    SELECT st_X(st_PointN($2,n)) AS t
> 	    FROM generate_series(1,st_NumPoints($2)) AS h(n)
> 	    UNION
> 	    SELECT st_Xmax($1) AS t
>             ) AS q ORDER BY q.t) AS xy
> ...bottom...
> dr_delay_value() is simply a look-up function that takes the Y value for a
> certain X.
>
> The thing is that eventually this fuction is failing on returning more
> specifically a 2-points linestring (that sould) and returns only a single
> point one. Now, I have prepared a "wrapper" PL/Pgsql function to keep track
> of what is passed to that function (perhaps that was the reason for the
> error. With that I'm pretty much sure that the arguments passed are fine,
> and still get the same error... Strangely, with my wrapper function keeping
> track of the arguments passed to the function I was able to try out to run
> the same request (that inside of the bigger function fails) separately, and
> guess what? is simply works!!
>
> I hope anyone may have a clue of what is going on. That's a very strange
> behavior, I would say.
>
> Regards,
>
> Rodrigo Sperb
>