| Re: converting VMS idl files to UNIX pvwave [message #473] |
Thu, 16 July 1992 07:47 |
thompson
Messages: 584
Registered: August 1991
|
Senior Member |
|
|
In article <1992Jul16.004709.31221@urz.unibas.ch>, hofer@urz.unibas.ch (Remo
Hofer) writes...
> In article <1992Jul11.034822.19367@bnlux1.bnl.gov>,
> jacobsen@bnlls1.nsls.bnl.gov (Chris Jacobsen) writes:
>> I had the same question. Mark Rivers of University Chicago/Brookhaven
>> Lab set things up using the XDR file format option. That allows
>> me to ship files back and forth between VMS and Unix
>
> Since XDR uses a description tag for every scalar data and not for a whole
> array, arrays written with XDR are often two times as big as without XDR.
> We ended up only wrtiting a stream of bytes to our data files using our own
> portable convertion routines to convert e.g. an array of floats into an array
> of bytes. We are transfering data (floats, longints, ints, bytes) between a VAX
> running VMS (little endian, VMS floats) and a Silicon Graphics Iris
> (big endian, IEEE floats) without any troubles this way.
I thought I might mention that IDL now supports converting floating point
numbers with the BYTEORDER function. The relevant calls would be for
converting from XDR (IEEE) formats to the host format:
BYTEORDER,DATA,/NTOHS ;Short integer
BYTEORDER,DATA,/NTOHL ;Long integer
BYTEORDER,DATA,/XDRTOF ;Floating point
BYTEORDER,DATA,/XDRTOD ;Double precision
BYTEORDER,DATA,/XDRTOF ;Complex
and from host to XDR (IEEE) formats
BYTEORDER,DATA,/HTONS ;Short integer
BYTEORDER,DATA,/HTONL ;Long integer
BYTEORDER,DATA,/FTOXDR ;Floating point
BYTEORDER,DATA,/DTOXDR ;Double precision
BYTEORDER,DATA,/FTOXDR ;Complex
That way you avoid the description tags. These calls are very fast.
> The bigger problem is the record formats on VMS.
> I'm not able to read a big file in stream format with pv~wave routines.
> (An error message saying that some buffer is too small is shown.)
> Converting the file to stream_lf format helps about this.
> Transfering a file written with pv~wave on the VAX in variable lenght record
> format to the Iris using ftp results in a mixture between actual data and
> record info on the unix side. And these are just two examples of troubles with
> VMS record orientated files.
If you use fixed length or stream records, rather than variable length, then it
is possible to access particular bytes in a file using
OPEN,/BLOCK Opens the file for block access (on VAX)
POINT_LUN Select starting byte to read
READU Read data
WRITEU Write data
This is more portable than ASSOC, or any of the other more record-associated
I/O calls. That way, on the VAX, then the actual file attributes are ignored.
The file has to be written so that the bytes follow one another without
interruption. Therefore, it won't work with variable length record files which
have extra embedded bytes describing the records, which is different between
VAXs and other computers.
I use the above commands to read and write FITS files, which can be transported
without change between different computers, in particular, VAX/VMS computers
and Sun workstations.
Bill Thompson
|
|
|
|
| Re: converting VMS idl files to UNIX pvwave [message #475 is a reply to message #473] |
Wed, 15 July 1992 15:47  |
hofer
Messages: 16
Registered: April 1991
|
Junior Member |
|
|
In article <1992Jul11.034822.19367@bnlux1.bnl.gov>, jacobsen@bnlls1.nsls.bnl.gov (Chris Jacobsen) writes:
> I had the same question. Mark Rivers of University Chicago/Brookhaven
> Lab set things up using the XDR file format option. That allows
> me to ship files back and forth between VMS and Unix
Since XDR uses a description tag for every scalar data and not for a whole
array, arrays written with XDR are often two times as big as without XDR.
We ended up only wrtiting a stream of bytes to our data files using our own
portable convertion routines to convert e.g. an array of floats into an array
of bytes. We are transfering data (floats, longints, ints, bytes) between a VAX
running VMS (little endian, VMS floats) and a Silicon Graphics Iris
(big endian, IEEE floats) without any troubles this way.
The bigger problem is the record formats on VMS.
I'm not able to read a big file in stream format with pv~wave routines.
(An error message saying that some buffer is too small is shown.)
Converting the file to stream_lf format helps about this.
Transfering a file written with pv~wave on the VAX in variable lenght record
format to the Iris using ftp results in a mixture between actual data and
record info on the unix side. And these are just two examples of troubles with
VMS record orientated files.
Remo Hofer
--
RFC822: <hofer@urz.unibas.ch> or <hofer%urz.unibas.ch@CERNVAX.BITNET>
X.400: S=hofer;OU=urz;O=unibas;P=SWITCH;A=ARCOM;C=CH
HEPNET/SPAN: CHGATE::YOGI::HOFER or 20579::48130::HOFER
|
|
|
|
|
|
|
|
| Re: converting VMS idl files to UNIX pvwave [message #490 is a reply to message #487] |
Thu, 09 July 1992 09:24 |
pan
Messages: 6
Registered: February 1992
|
Junior Member |
|
|
In article <1992Jul7.101543.556@ualr.edu> ptfinley@acs.harding.edu writes:
> I have a data file that was created by an idl procedure on a VAX
> system. I need to read the data in a pvwave procedure on a
> UNIX system. The problem seems to be that the VAX uses records
> and has a header at the top of the file that explains the records
>
> My UNIX system merely reads the first byte with the assumption that
> this is the desired data. If you have done this before, I would
> appreciate any hits you could give. I need to remove the header,
> but I also need to know the record lengths. THANKS.
>
> *****************
> * Paul Finley, Harding University, Searcy, AR
> *
> * PTFINLEY at ACS.HARDING.EDU
> * or
> * PFINLEY at SUN1.HU.EDU
> ***************
I think VAX record file can be converted to stream (or unformatted) file
using IDL on VAX. After the conversion, you can send it over to unix
system then read it from unix IDL or PV-WAVE. However, one tricky
problem here is that you may have to do byte-swapping on the conversion
file (if integer format), or format-conversion (tricky too) if floating
point format. Depends on the machine you are using (DEC5000 or IBM6000
and others), the floating point conversion will also be different.
This sketchs how I coped with this problem. Maybe there is another
better approach.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
