| Comparing tabulated functions [message #57841] |
Thu, 03 January 2008 07:47  |
jameskuyper
Messages: 79
Registered: October 2007
|
Member |
|
|
I've got two functions of time, each of which is tabulated at
thousands of different monotonically increasing time values; in the
worst case, there may be as many as 14,400 time values. I want to
estimate the difference between these functions. Unfortunately, the
two functions will not, in general, be tabulated at the same time
values. A simple approach that works fairly well is to use spline
interpolation to interpolate both of them to a common grid, and then
difference the interpolated the values. However, occasionally one or
the other of the two functions have data gaps. For any time value that
is more than 1 second away from the nearest tabulated time for either
function, I want to fill in the difference array with !VALUES.D_NAN. I
believe this will cause the PLOT command to skip that point - if not,
I need to find some other approach that will have this effect.
I can write simple, highly efficient C code that does exactly what I
want; translating that code into IDL would make heavy use of loops,
and therefore wouldn't be very efficient. I can write simple IDL code
that takes no advantage of the fact that the arrays are monotonically
increasing, calculating the difference of every point on the output
grid from every point on either of the two input tables, and finding
the minimum, but that seems extremely inefficient (and a memory hog!).
Is there a simpler way to do this?
|
|
|
|
| Re: Comparing tabulated functions [message #57934 is a reply to message #57841] |
Sat, 05 January 2008 15:42  |
jameskuyper
Messages: 79
Registered: October 2007
|
Member |
|
|
Lasse Clausen wrote:
> Hi,
>
> On 3 Jan, 15:47, jameskuy...@verizon.net wrote:
>> I can write simple, highly efficient C code that does exactly what I
>> want;
>
> excuse my heretic question: why would you want to use IDL, then?
Convenience. The code to bring the data into an IDL context already
exists, and I wanted to use the results as data for IDL plotting
commands. Therefore, to bring C into this, I would have had to interface
between C and IDL twice, which is moderately complicated, no matter how
you do it.
I was pretty sure that there was an IDL feature whose name I couldn't
quite remember that would allow me to deal with this with relative
efficiency and simplicity in pure IDL. I was right - VALUE_LOCATE is
what I was looking for.
|
|
|
|
| Re: Comparing tabulated functions [message #57968 is a reply to message #57841] |
Fri, 04 January 2008 09:35 |
lasse
Messages: 48
Registered: February 2007
|
Member |
|
|
Hi,
On 3 Jan, 15:47, jameskuy...@verizon.net wrote:
> I can write simple, highly efficient C code that does exactly what I
> want;
excuse my heretic question: why would you want to use IDL, then?
Regards
Lasse Clausen
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
