[ENH] Create Class widget #1766
Conversation
Current coverage is 89.18% (diff: 100%)@@ master #1766 diff @@
==========================================
Files 85 85
Lines 9027 9063 +36
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 8039 8083 +44
+ Misses 988 980 -8
Partials 0 0
|
janezd
force-pushed
the
create-class
branch
5 times, most recently
from
66f3a0a
to
b398883
Compare
janezd
force-pushed
the
create-class
branch
5 times, most recently
from
9af2d65
to
4322195
Compare
| def map_by_substring(a, patterns, case_sensitive, at_beginning): | ||
| def map_by_substring(a, patterns, case_sensitive, match_beginning): | ||
| """ | ||
| Map values in a using a list of patterns. The patterns are considered in |
There was a problem hiding this comment.
Map values in a using a list of patterns.
- What patterns?
|
I like this widget a lot, but GUI totally confuses me. The + symbol is very small and renders as cross on Win, instance count odd, widget is a bit long... I'm not sure, it just doesn't lie well with me. As for functionality, I'd like to have sth like Other, Rest or *... Something that says 'everything else belongs to this class'. Also as for pattern, how about Regex? If not, I'd have a tooltip for 'Pattern' giving an example of use. Will be featured in docs, too. I'd add Report and have the two bottom buttons as in Feature Constructor. |
|
Note to Besides what Ajda wrote above:
|
|
It worked with spectral data and Files widget. |
|
Could you also include a help file for the widget? Even if it is just a stub. |
In line with my widely known and well established tradition of impeccable pull requests and thoroughly tested code, I am presenting a new widget that creates a class attribute with values based on substrings of the chosen string or discrete attribute. The widget is needed for infraorange and presumably also in bioinformatics.
The widget has zero pylint warnings, was assigned a straight A by radon, and has a 100 % coverage with unit tests of the highest quality. The latter rely on simulated user interaction with the widget, and restrict themselves from testing the implementational details except where appropriate. The abundance of documentation strings explaining even the smallest nuances of the private methods - but without going into unnecessary details - will make the future maintenance of the widget pure joy.
This whole grand endeavour is topped by a carefully crafted icon that complements the widget's amazing user interface.
A minor shortcoming that may need to be mentioned is that I can't make the widget properly shrink vertically when lines are removed. Not for lack of trying. Any help appreciated.
I have not written any documentation for the widget catalog so far. This will wait for any suggestions and for @ajdapretnar's help with screenshots.
The code of the widget contains two general classes derived from
TransformandLookup-- but perhaps not general enough to pollute the namespace of thetransformationmodule.The pull request includes a fix for
Orange.preprocess.transformation.Lookup, which probably nobody used since it was plain broken. This commit could belong to another pull request, but processing/merging chains of pull requests tends to take too long in my experience. Being broken and useless, the class has two relatives of the same name in other modules, one of which does exactly the same thing as the originalLookupdid not do.@markotoplak, can you test this on files with hyperspectral data? After this glorious presentation, let's hope the widget doesn't crash on the very first use. ;)
Includes