[Fwd: From RSI: problem reading non-finite data in IDL 6.1]

From: Tony Spadafora (ALSpadafora@lbl.gov)
Date: Fri Oct 01 2004 - 18:13:34 PDT

  • Next message: Tony Spadafora: "Proposed SCP collab meeting Oct 26"

    Begin forwarded message:

    > From: RSI E-mail Support <support@rsinc.com>
    > Date: October 1, 2004 4:42:33 PM PDT
    > To: "'LARadding@lbl.gov'" <LARadding@lbl.gov>
    > Subject: From RSI: problem reading non-finite data in IDL 6.1
    >
    >
    > Dear Elizabeth Radding:
    >
    > This email is to alert IDL users to an error introduced while making
    > improvements to IDL 6.1. A patch will be available within the next few
    > weeks to fix this problem.
    >
    > Problem description:
    >
    > This problem affects the READ set of commands (READ, READF, and READS)
    > when
    > a natural width floating point format code (F0,
    > D0, E0 and G0) is specified via the FORMAT keyword. In this
    > situation, if
    > IDL encounters a non-finite value such as "NaN" or "Inf", it will not
    > advance the input buffer past it. Hence, any remaining data items for
    > the
    > record also receive the same non-finite value. Because no error is
    > thrown,
    > a user may not be aware that the data values following NaN values in a
    > record are incorrectly reported as NaN.
    >
    > Solution:
    > To avoid this problem until the patch is available from RSI, avoid
    > using
    > natural width format codes when reading floating point data. The ID
    > number
    > for this issue is CR38238.
    >
    > Reproduce case:
    > Save the following data into "test_nan.dat"
    > 342.34 200.00 1
    > 325.61 201.02 1
    > 326.78 NaN 0
    > 330.21 203.05 1
    >
    > Executing the following code:
    > infile='test_nan.dat'
    > openr,lun,infile,/get_lun
    > x=0. & y=0. & z=0. & i=0
    > readf,lun,x,y,z, format='(F0,F0,F0)'
    > readf,lun,x,y,z, format='(F0,F0,F0)'
    > readf,lun,x,y,z, format='(F0,F0,F0)'
    > print,x,y,z
    >
    > results in the following incorrect behavior:
    > 326.780 NaN NaN
    >
    > Previous IDL versions correctly report:
    > 326.780 NaN 0.000000
    >
    > In addition to providing a patch for this problem, we will also
    > evaluate
    > other, minor issues to include in the patch. We apologize for any
    > inconvenience. If you have questions regarding IDL 6.1, please contact
    > Technical Support at support@rsinc.com or 303-413-3920.
    >
    > Sincerely,
    >
    > Technical Support
    > Research Systems, Inc.
    > 303-413-3920
    > support@rsinc.com
    >
    >

    This archive was generated by hypermail 2.1.4 : Fri Oct 01 2004 - 18:13:58 PDT