| Avoiding multiple or complex WHEREs? [message #75224] |
Mon, 21 February 2011 13:31  |
cgguido
Messages: 195
Registered: August 2005
|
Senior Member |
|
|
;Say I have an array of point IDs and times where p[0,*] are IDs and
;p[1,*] are times, something like:
p=[[1,1],[1,2],[2,3]]
;Typically, the array has many points, many times and many gaps
;I want to be able to get a subset of 'p' that contains a range
;[id0:idn] or a list [id0,id2,id7] of IDs and Ts the most efficient
;way possible.
;Right now, I build an array 'i' with one FOR loop where i[id,t] is
the row index to 'p'
;with the correct id and t. If that combinations does not exist, then
;i[id,t]=NaN. So:
i[0,0]=NaN
i[2,0]=NaN
i[1,*]=[NaN,0,1,NaN]
;So if I want all instances of particle 1 in array p I do
i1=i[1,where(i[1,*] ge 0)]
p1=p[*,i1]
;This gets handy when I want particles 1,3,4 and times [23:100],
;because I don't have to do where(id eq 1 OR id eq 3 OR id EQ 4....
etc.)
i_cool=i[1,where(i[[1,3,4],[23:100]] ge 0)]
p_cool=p[*,i_cool]
;Now, I am betting that a histogram/histobin trick might be even
;better... Any ideas?
;PS: This does not deal nicely with IDs or Ts that are bigger than the
;biggest contained in 'p'
|
|
|
|
| Re: Avoiding multiple or complex WHEREs? [message #75301 is a reply to message #75224] |
Tue, 22 February 2011 12:02  |
cgguido
Messages: 195
Registered: August 2005
|
Senior Member |
|
|
On Feb 22, 10:29 am, Paulo Penteado <pp.pente...@gmail.com> wrote:
> On Feb 21, 6:55 pm, Gianguido Cianci <gianguido.cia...@gmail.com>
> wrote:
>
>> Yeah :-/ I guess I am trying to build something like a HASH with two
>> keys for each value... but I could be going about this completely
>> backwards!
>
> It depends on how you are going to use it, but this problems suggests
> to me making a class that overloaded the brackets, to search for what
> you want, and return !null if not found. Internally it could use
> hashes, lists or arrays (of values or pointer arrays). The better
> choice depends on things like the number of points and whether they
> form a grid in ID, whether Ts are integers (I am guessing that
> particle IDs are). That way it would be usable doing things like
>
> p_cool=p[[1,3,4],23:100]
Wow. I did a little object oriented programming a long time ago, in
java, but never with IDL. DIdn't realize that you could overload even
the brackets operators. That's really cool! Not sure exactly how this
would work though, seen as in reality 'p' has a few more columns
(x,y,z,mass,radius,T,ID) so I would want to overload the operator for
the last two columns but still be able to use plot, p[0,*], p[1,*] to
plot al the positions. So for columns 0:4, the brackets should not be
overloaded. And since you mentioned it, ID and T are both integers,
not all ids may be present at all times, there could be 10k to 100k ID-
Time pairs in each set.
Thanks for your help Paulo!
Gianguido
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
