Robustness and reliability of Sherpa 2-D image fitting have been
Support for dummy instruments and data sets has been added.
Check details in ahelp instrument.
Users can now set contours levels in contour plots generated
by cplot via a new variable in
the Sherpa state object, sherpa.con_levs; see
ahelp sherpa for details. This
change applies to cplotonly; it has no effect on confidence plots generated
with region-projection and region-uncertainty.
Variable - sherpa.deleteframes - has been added to
the state object. It allows for deleting or creating ds9 frames
with IMAGE command. See ahelp sherpa for details.
Support for multi-dimensional data on non-uniform grid has been added.
Changes to Models
Radial profile analysis have been improved. The two instrument models
fpsf1d and tpsf1d, have now additional parameter, radial,
which needs to be set to 1 for the radial profile analysis.
The order of parameters in beta2d model has been
changed. The alpha parameter is now last.
The parameter convTyp has been renamed to
fft in the following models: fpsf1d,
fpsf2d, tpsf1d, tpsf2d,
and rsp2d. If fft = 1,
use Fast Fourier Transforms in convolution; if fft =
0, don't (i.e. use the sliding cell method).
The default integration setting has been changed from ON to
OFF for all 2D models exceptconst2d, delta2d, polynom2d. For
typical 2D analyses in image coordinates, the difference
between ON and OFF is that ON integrates the model over the
pixel, thus takes into account some model curvature within
that pixel, while OFF assumes the value at the center of the
pixel and can lead to faster analyses.
Changes to Sherpa/S-Lang Module
Several functions have been added to the Sherpa/S-Lang module:
returns a string showing the current analysis setting for
the specified dataset.
sets the analysis for the specified dataset.
get_coord - returns
a string showing the current coordinate setting for the
set_coord - sets
the coordinate for the specified dataset.
The Sherpa/S-Lang module now links to the CHIPS library, instead of
importing the chips module.
CAST command has been disabled. Sherpa now uses double precision
variables internally. Numeric output may have changed slightly from
the one in the previous releases.
PLOTX command has been disabled. Plots are done using the ANALYSIS
A number of fixes have been made that greatly improve the robustness
and reliability of image fitting in Sherpa.
image fit and image model do not have any offsets.
Coordinate system for both source and background 2-D data is now
updated with coord command.
Guessing the initial parameters for box2d model is now
correct when "COORD PHYSICAL" has been set.
const2d model (c0 parameter
> 0) is now correct for "COORD WCS"
Parameter guessing takes the exposure map into account for 2D data.
The WRITE MODEL
command correctly applies the filter for 2D images.
load_bimage is retaining WCS
The fpsf1d instrument model has been fixed.
The treatment of read quality
flags in grouped PHA data is correct.
Models are now correctly evaluated with filtered data that contain
Filtering out a single bin (e.g. with "ignore bin 1") is
RESET FILTER now resets both source and background filters.
Problems with applying energy filters to 1D data has been fixed.
Setting an instrument model does not change the analysis setting.
Reading MDL files has been corrected.
Sherpa will display an error message if data directory does not exist.
Sherpa models are not limited to the float domain.
WRITE command is now working properly for datasets of > 2 dimensionality.
Strange parser behavior due to a failed S-Lang assignment is fixed.
The function get_axes should work properly with
1-D binned dataspace.
The function get_eflux has been fixed.
Corrected the set_stackexpr function to return 1 (indicating
success) only for valid source expressions.
A number of functions (e.g. sherpa_eval,
could cause a
"Stack Underflow" error in the S-Lang interpreter. This has now been fixed.
Functions get_dcounts_sum and get_dcounts_sum work now
correctly with background subtracted data.
run_fit function is now allowed with sigma-rejection
Files in $ASCDS_INSTALL/contrib/doc/xml are now included in
Fixed bug where 'line' clipped incorrectly in log scale
Fixed problem with crashing when printing to long filenames
Fixed XPA error that occurred when exiting from batch mode under Mac OSX
No longer resets text field entries to the first
character of a string when the focus leaves the text entry widget (allows
end of long file names to stay in view)
Now displays current working directory info in the status message frame
Warning message no longer appears
when an "Increment/Decrement Scale" dialog is invoked (affected
Redhat 6.2, Redhat 8, and Mac OSX platforms)
Warning messages no longer appear when
a parameter file is opened with peg.par's showhidden parameter set to no,
and the user subsequently checks "Show Hidden Parameters" in the
"Peg Preferences" dialog
(affected Solaris 8 and Alpha platforms)
Histogram functionality now works on Alpha platform
Using the "Expand Column" dialog no longer generates a spurious cell-edit
request (affected Redhat 6.2, Redhat 8, and Mac OSX platforms)
'Prism Preferences' window now has a vertical scroll bar
Fixed problem with long custom analysis menus running off the desktop
CIAO Setup Scripts
The CIAO setup scripts (ciao.*sh) now provide enhanced support for
contributed and custom software installed in
$ASCDS_INSTALL/contrib. If they exist, the subdirectories
bin, bin/interpreted, param, and
lib of $ASCDS_INSTALL/contrib will be treated identically
to their counterparts in $ASCDS_INSTALL. This means that
command-line scripts, S-Lang scripts, script parameter files, and
libraries installed in the appropriate place in
$ASCDS_INSTALL/contrib will be automatically "found" by CIAO.
The CIAO contributed scripts package (CIAO_scripts.tar)
has been reorganized to make use of this new functionality, allowing for
near-seamless integration of the contributed scripts with CIAO.
See the scripts page for more information.
Resuming a CIAO analysis session via "ciaostart <savefile>" or
"filtwin <savefile>" now works under Redhat 6.2, Redhat 8, and
Now defaults to fixed-width font for improved output formatting
Command-line options now handled correctly under Redhat 6.2 and RedHat 8
No longer corrupts files (e.g. XMM spectra) that contain region blocks
that are "attached" to the data inside another block when opening them
Removed extraneous text in WCS info when listing column information
Now warns when user tries to bin more than 32767 counts in to one pixel,
which causes pixel value to "wrap" into negative values
Fixed problems that appeared with multiple block pointers to same block
open for writing
Fixed bug that appeared when binning and region filtering at the same
time with different binning factor in x and y: Region got scaled by the
bin scale of x only
dmlist/dmcopy can now handle images of data type dmBYTE
Added warning message about trying create an image larger than 75MB;
image will still be created unless user aborts
Fixed bug that prevented using "exclude" with bit filters
dmlist now prints warning and truncates large arrays in 'data' mode
Restored ability (lost in CIAO 3.0) to filter with GTI files that do
not contain a data subspace (e.g. those created by FTOOLS)
Restored behavior (lost in CIAO 3.0) of making the default GTI subspace
for a GTI block be the block itself
Fixed floating point rounding error that could cause incorrect sizes
for filtered images
Fixed problem where bad binning syntax sometimes failed to produce an
Fixed problem where using the "bounds" syntax to create a region which is
the bounding box of another region caused the new region to be bounded
by a 0 - 32768 box; region is now bounded by the
limits (TLMIN/TLMAX) of the vector to which the region is applied
dmGetScalars can now handle any number of rows
For the case of binning a file with a filter that results in 0 rows, the DM
now returns an image of all 0s. Previously, dmImageDataGetSubarray() was
returning the pixel array back to the user without setting any values,
which can appear to be a random array of values if the array is not
Fixed bug where #row filtering
would not work in conjunction with dmGetScalars() or another function
that requests a specific number of rows when the rows passing the
#row filter were sparse (separated by > ~1500 rows)
Now possible to set the value of a
vector component using the dmSetScalar functions. In CIAO 3.0 and 3.0.1,
the DM would
behave as if dmSetVector() was called, setting the first component of
the vector to the argument of dmSetScalar and setting any following
components to whatever resided in memory after the first value.
Fixed bug where mkrmf generated a one-bin shift if user-specified energy
lower bound was beyond the FEF energy lower limit
Previously, VARMM separated groups in PHA files by the grouping
flag (1, -1, etc...) as well as the quality flag. If the
quality flag was not zero then the channel was set as a new
grouping is independent of the quality, and
the group's quality is set to the highest quality flag value in
the group. (This makes Sherpa's treatment of quality flags more
consistent with XSPEC's.)
VARMM no longer clears the stack when doing a readXXX. This caused
problems when attempting to push values on the stack
and evaluating the readXXX afterwards.
The readImage function was reading in pixels flagged as type
BYTE as signed characters. They are now read in as unsigned