| Re: inexplicable LONG() - behaviour [message #2871 is a reply to message #2870] |
Wed, 07 September 1994 02:19  |
steinhh
Messages: 260
Registered: June 1994
|
Senior Member |
|
|
In article <34hrvh$6sq@jurpool0.rz.uni-frankfurt.de>, frank@chaos.uni-frankfurt.dbp.de (Frank Hoffsuemmer) writes:
|> Hello,
|>
|> I'm using IDL 3.1.1 (no update in sight :-( ) under HP-UX.
|> And there are some strange things happening....
|> Of course, these are just things that I understand wrong :), so could someone
|> please explain this behaviour:
|>
|>
|> IDL. Version 3.1.1 (hp-ux hp_pa).
|> Copyright 1989-1993, Research Systems, Inc.
|> All rights reserved. Unauthorized reproduction prohibited.
|> Installation number: 3063.
|> Licensed for use by: Johann Wolfgang Goethe-Universitaet, HRZ
|>
|> X-IDL> print, long(1231231434.1)
|> 1231231488
|>
|> The last two digits differ quite a bite. O.k. the number is too long,
|> but why isn't there an error-message like in this example (same number but
|> first digit):
Because the value 1231231434.1 isn't possible to store in a FLOAT
(not enough precision). (It's actually stored as 1231231488!)
The error occurs in converting the textual digits into a float,
which is not normally considered an error, as long as it's the
float precision causing it.
Try print,long(1231231434.1d)
Or, print,'$(f12.1)',1231231434.1
|>
|> X-IDL> print, long(2231231434.1)
|> % Program caused arithmetic error: Floating overflow
|> % Detected at $MAIN$ (LONG).
|> 2139095040
|>
|> there still is an result, and the result is still wrong, but at least there's
|> a message indicating that something went wrong.
|>
Such a large number can be stored in a FLOAT (because of the FLOAT-ing
point exponent notation), but not in a LONG (just not enough bits).
Thus, a significant error occurs, and it's reported.
|> Furthermore, I programmed a little routine, that is supposed to return
|> a rounded value of it's argument (included at the end of this article).
|> The argument can be integer, long, double, and the result is always long. e.g.
|>
|> X-IDL> PRINT, ROUNDUP(12345.9867)
|> 13000
|> X-IDL> PRINT, ROUNDUP(999.67)
|> 1000
|> X-IDL> PRINT, ROUNDUP(53645)
|> 54000
|>
|> - just to give you an impression.
|> Using this function I have the following problem:
|>
|> X-IDL> PRINT, ROUNDUP(101.1)
|> 109
Your problem is that what's calculated in your procedure *isn't* the
value double(110.0), its (change of the print lines to
PRINT, '$(A,f19.15)',"DOUBLE: ", DOUBLE(result)
PRINT, "LONG : ", LONG(result+0.0001)
And voila:
RESULT: 110.00000
DOUBLE: 109.999999999999970
LONG : 110
Now, it's the DOUBLE that's wrong... :-)
It's all correct, if you allow for the finite-precision nature of
things.
Stein Vidar
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
