| Need integer consts eq long not short [message #3100] |
Wed, 02 November 1994 12:32  |
sgs
Messages: 2
Registered: November 1994
|
Junior Member |
|
|
I need pvwave to have long (4 byte) constant integers by default rather
than short (2 byte). Is there a way to force this behavior with a
system variable, or something?
Short integers cause problems like this:
for i=0,n do begin ... ; buggy, i is a short integer
for i=0L,n do begin ... ; correct, i is a long integer
Pvwave creates i as a short integer, because the constant 0 is, by default,
short. However, when (long) n is bigger than 32767, the loop only loops
to (n mod 32767).
(1) Unfortunately, these loops often produce reasonable, but incorrect
answers.
Many of the std/* and user/* routines provided with pvwave have this
bug. (grep for lines with "for i=0", and any lines with integer
constants). (Watch out! They work with your small test case, but fail
with your real, large dataset, see (1).)
Steve Spray sgs@lmsc.lockheed.com
|
|
|
|
| Re: Need integer consts eq long not short [message #3144 is a reply to message #3100] |
Thu, 10 November 1994 06:33  |
landers
Messages: 45
Registered: May 1993
|
Member |
|
|
In article <39qs0h$g7@ovid.dra.hmg.gb>, black@signal.dra.hmg.gb (John Black) writes:
|>
|> While we're on the subject of bugs due to short integers rather than long
|> integers, I might point out another variation of bugs due to this.
|>
|> This is using the FIX function as opposed to LONG. A example of this is in the
|> PV-WAVE routine SUM. The routine works so long as the output array is not over
|> 32767 elements. I think this particular routine only exists in PV-WAVE - the
|> IDL routine that performs the equivalent operation is TOTAL (note the IDL
|> version of total not the PV-WAVE one). This particular bug is still in SUM in
|> the version of PV-WAVE that we have which is 4.20
|>
|> Note also that AVG calls SUM, so there's effectivly a bug in AVG too.
|>
|> John Black
PV-WAVE has bot TOTAL and SUM procedures. TOTAL just totals the contents of
an array, and is an 'intrinsic' function (it's compiled in - there's no
total.pro source). SUM is a User's Lib routine that will sum values from a
particular array index.
The bug you mention in SUM has been fixed for version 5.00, as well as for the
PC version, PV-WAVE Personal Edition.
--
Dave
|
|
|
|
| Re: Need integer consts eq long not short [message #3146 is a reply to message #3100] |
Wed, 09 November 1994 08:05 |
black
Messages: 39
Registered: August 1992
|
Member |
|
|
While we're on the subject of bugs due to short integers rather than long
integers, I might point out another variation of bugs due to this.
This is using the FIX function as opposed to LONG. A example of this is in the
PV-WAVE routine SUM. The routine works so long as the output array is not over
32767 elements. I think this particular routine only exists in PV-WAVE - the
IDL routine that performs the equivalent operation is TOTAL (note the IDL
version of total not the PV-WAVE one). This particular bug is still in SUM in
the version of PV-WAVE that we have which is 4.20
Note also that AVG calls SUM, so there's effectivly a bug in AVG too.
John Black
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
