Converting list values to an attribute depending on other list values

Hi All,

In 2016.0, I have a list like this:

_list{0}.class = 1

_list{1}.class = 2

_list{2}.class = 3

_list{3}.class = 5

_list{4}.class = 2

_list{5}.class = 1

I need to know how many of any given class there are with the feature.

ListHistogrammer does this nicely:

_histo{0}.value = 1

_histo{0}.count = 2

_histo{1}.value = 2

_histo{1}.count = 2

_histo{2}.value = 3

_histo{2}.count = 1

_histo{3}.value = 5

_histo{3}.count = 1

My problem then is turning those into attributes simply. I don't want to have to explode or deconstruct my feature further (this workspace is already too large and slow).

In this example, I'd like my output feature to have attributes like these:

first_class = 2

other_thing = 2

third_item = 1

fourth_item = <missing>

final = 1

Any ideas how to do this? I could likely do it with a PythonCaller, but surely there's a better way?

Page 1 / 1

If you want to flatten the histogram as if it were a key/value pair,

in your histogram example you would end up with

1: 2

2: 2

3: 1

5: 1

then creating a custom transformer loop and using an attributeCreator with the new attribute as @Value(_histo{@Value(i)}.value) and the Attribute Value as @Value(_histo{@Value(i).count) should work. Where i is the loop iteration from 0 to length of list -1

If you want the attribute to be named first_class, fourth_item, etc, then python is the way to go.

Upon further investigation, I'm not even sure this is possible with Python! I can't find any way to access a full list - this question from 2012 suggests the "new" API doesn't have one - https://knowledge.safe.com/questions/5463/python-and-lists.html

Is it possible?

feature.getAttribute('_histogram{}.value') will return a list as will feature.getAttribute('_histogram{}.count').

If necessary you can zip them together, though I have rarely needed to do that.

I cannot understand the naming rule for the destination attribute ('first_class', 'other_thing' ... 'final'?), but if the goal was just to map the counts to new attributes with any names corresponding to the values (1, 2, ... 5), this Python script example might help you.

# PythonCaller Script Example

def processFeature(feature):
    valueToAttr = {
        '1': 'first',
        '2': 'second',
        '3': 'third',
        '4': 'fourth',
        '5': 'fifth',
    }
    
    valueToCount = {}
    values = feature.getAttribute('_histo{}.value')
    counts = feature.getAttribute('_histo{}.count')
    if values and counts:
        for v, c in zip(values, counts):
            valueToCount(v] = c
            
        for v in valueToAttr.keys():
            if v in valueToCount:
                feature.setAttribute(valueToAttr�v], valueToCount�v])

# PythonCaller Script Example

def processFeature(feature):
    valueToAttr = {
        '1': 'first',
        '2': 'second',
        '3': 'third',
        '4': 'fourth',
        '5': 'fifth',
    }
    
    valueToCount = {}
    values = feature.getAttribute('_histo{}.value')
    counts = feature.getAttribute('_histo{}.count')
    if values and counts:
        for v, c in zip(values, counts):
            valueToCount(v] = c
            
        for v in valueToAttr.keys():
            if v in valueToCount:
                feature.setAttribute(valueToAttr�v], valueToCount�v])

Thanks, that did the trick nicely, I couldn't find anything how to access lists with Python, hence my problem. Relating to the naming convention, I was trying to express that the attribute names would be arbitrary, but that's easily resolved in your version by changing the dict values.

Reply

Community Stats

Reply

Community Stats

Sign up

An FME Account is required to contribute

Login to the community

An FME Account is required to contribute

Scanning file for viruses.

This file cannot be downloaded