also skips __getitem__ coverage update if the full image data range has
been calculated.

imagewrapper module.

between expansions.

no sense, I know.

data is read. But the dataRange property will return cal_min/cal_max
if no data has been read. Some documentation added too.

comparisons to None now testing for isnan.

ImageWrapper now maintains separate coverages for each volume (in a 4D
image, or slice in a 3D image), and adjusts them individually. Lots of
crazy code which seems to work ok, but has not been tested or
documented. Most of the complex stuff has been turned into standalone
functions, so can be unit-tested.

arg. When a Notifier listener is registered, it can be specified to run
on async.idle (to avoid threading issues).

As an aside, I have just realised that there is a naming collision
between Notifier.notify and HasProperties.notify, for anything which
sub-classes both. I don't think anything does, but this will probably
bite me in the future.

files. Log messages added to memoize functions.

HasProperties type).

initialisation. Image.name property turned into a regular (writeable)
attribute.

changes. Bugfix in Image.__getitem__ - was potentially returning data
with trailing dimensions of size 1.

function.

future, but it's distracting me for the time being, so out it goes.

completely controlled by calling code.  The partial data range
calculation behaviour should be decided by FSLeyes (or whatever
is working with the Image object).

good measure. More work to be done.

hopefully will handle writing too). Read sort of working. Major
disruptive changes to Image class. Things breaking bad.

and allows caller to specify which arguments to use.

things.

settings).

button.

duplicating functionality provided by the props.callqueue module.

we're running locally or over ssh.

error.

habit of calling it mutliple times from different threads. Some updates
to AtlasDescription class - am intending to allow users to select
atlas resolution.