Issue with calling transform() on linear and glog

[dmoccia]

Issue with calling transform() on linear and glog

[eyurtsev]

thanks for reporting this looks like something is passing an extra 'd' argument down the road

in the meantime you can define your own custom transformations if you need a workaround

http://eyurtsev.github.io/FlowCytometryTools/gallery.html#custom-transformations

[dmoccia]

No problem on the report, nice looking library. I will try the custom transforms out.

[dmoccia]

I believe I figured out why this issue is happening. In your FCMeasurement.tranform method you have the following code

if isinstance(transform, Transformation): transformer = transform else: if auto_range: # determine transformation range if 'd' in kwargs: warnings.warn( 'Encountered both auto_range=True and user-specified range value in ' 'parameter d.\n Range value specified in parameter d is used.') else: print "in else" channel_meta = self.channels # the -1 below because the channel numbers begin from 1 instead of 0 # (this is fragile code) ranges = [float(r['$PnR']) for i, r in channel_meta.iterrows() if self.channel_names[i - 1] in channels] if not np.allclose(ranges, ranges[0]): raise Exception("""Not all specified channels have the same data range, therefore they cannot be transformed together.\n HINT: Try transforming one channel at a time. You'll need to provide the name of the channel in the transform.""") kwargs['d'] = np.log10(ranges[0])

Specifically the "isinstance" check is not working and therefore everything (linear, hlog, tlog, etc) is falling through to the final line kwargs['d'] = np.log10(ranges[0])

In this case d will always be
'd': 5.4185399219516617

and therefore for hlog, tlog this is not an issue (except if you want to use another d), and linear and glog it is an issue since neither take d as an argument. I do not use isinstance that often unless referring to built in types so I am not familiar with how to properly use it in a custom class. Probably not the most elegant solution but perhaps a simple function in the transforms.py code which allows you to check from a list?

Let me know if you need any more info

[eyurtsev]

Thanks for looking into this... This commit should fix the issue:

44e8352