- integer dtype images now have integer-aligned bins, with bin number
determined by a target value
- step size is automatically chosen based on a target image size
- bins and step arguments have default values 'auto'
When an ImageItem's data has an integer dtype, this patch ensures that
each bin of the LUT histogram contains the same number of integer
values, in order to avoid "spikes" in the histogram that are merely due
to some bins covering more integer values than others.
This commit needs testing (it was rebased from an old commit).
- Essentially a graphical interface to dict; all items have text and value
- Assigns previously-selected text after list is cleared and repopulated
- Get, set current value
This merge converts all absolute imports into relative imports throughout pyqtgraph.
This allows pyqtgraph to be embedded as a sub-package of larger packages.
- CSV headers include data names, if available
- Exporter correctly handles items with no data
- pg.plot() avoids creating empty data item
- removed call to reduce() from exporter; not available in python 3
- Gave .name() methods to PlotDataItem, PlotCurveItem, and ScatterPlotItem
* __init__.py now contains latest release version string
* installing from git checkout that does not correspond to a release commit
will result in a more descriptive version string
New Profiler class from Antony Lee:
* Simpler API (with backward compatibility)
* Individual Profilers can be enabled by environmental variable PYQTGRAPHPROFILE
* Loads only data that is currently visible
* Downsamples to avoid plotting too many samples
* Loads data in chunks to limit memory usage during downsampling
A new function profiling system is implemented. Most importantly, this
allows one to profile various internal functions directly by setting the
`PYQTGRAPHPROFILE` environment variable to a comma separated list of
function and method names, e.g.
PYQTGRAPHPROFILE=functions.makeARGB,ImageItem.render \
python -mexamples
Specifically, items in `PYQTGRAPHPROFILE` must be of the form
`classname.methodname` or `dotted_module_name.functionname`, with the
initial "pyqtgraph." stripped from the dotted module name.
Moreover, the overhead of inactive profilers has been kept minimal: an
introspective check of the caller's name (only if `PYQTGRAPHPROFILE` is
set) and a trivial function (not method) call per profiler call.
The new profilers rely on `sys._getframe` to find the caller's name,
although the previous system (passing the caller's name explicitely)
could certainly have been kept instead.
Finally the API of profilers has been changed: register a
profiling point simply by calling the profiler, and profilers are
automatically flushed on garbage collection. See the docstring of
`pyqtgraph.debug.Profiler` for more details.
