| Grumbling about setting double precision [message #78788] |
Tue, 20 December 2011 08:58  |
wlandsman
Messages: 743
Registered: June 2000
|
Senior Member |
|
|
I am converting a Python program to IDL which includes several hundred double precision data values. In Python the data values are written as follows:
a = [2721.64363164731, -1615.73035059635]
As described by David Fanning in http://www.idlcoyote.com/math_tips/double.html to make these double precision in IDL, it is not enough to convert the vector to double
IDL> print,double(a),f='(2f18.10)'
2721.6435546875 -1615.7303466797
and it is not enough to put a "d" at the end of one of the values
IDL> a = [2721.64363164731d, -1615.73035059635]
IDL> print,double(a),f='(2f18.10)'
2721.6436316473 -1615.7303466797
Instead, I must spend an hour in my editor putting a "d" at the end of every single data value.
But it would sure be nice to be able to somehow tell the IDL compiler to interpret numeric values as double precision. --Wayne
|
|
|
|
| Re: Grumbling about setting double precision [message #78826 is a reply to message #78788] |
Thu, 22 December 2011 12:08  |
Karl[1]
Messages: 79
Registered: October 2005
|
Member |
|
|
Editing: A lot of times, constant arrays like this come from someplace else and someone wrote a script or program to generate them. Such a script or program can be modified to add the 'd'. If that is not feasible, there are plenty of editors that support keyboard macros which can be run repeatedly for a given number of times. Still tedious but not as bad.
Regular expressions are also possible. In Notepad++:
Find: ([0-9]*[\.]*[0-9]*)([,\]])
Replace: \1d\2
Press Replace All and you're done. If that make you nervous, you can restrict the range of Replace All by selecting the text and checking the "In selection" box.
Memory: It is probably a mistake to default to double. It would take up more space and cause other problems. Variables that are created as a result of this array being part of the calculation would be promoted unnoticed to double as well, increasing memory usage even more. And the change in precision may affect results.
I do like the
b=[24.5d,9999.9]
approach. It expresses intent and doesn't seem that confusing (to me). It isn't perfectly backwards-compatible, but if someone wrote the above, they probably wanted a double array anyway.
It should be easy to do in IDL. IDL needs to analyze the first array element in order to decide what sort of array to make; int, string, or float. It would surely know if the literal is double or float at that point.
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
