@svteshipley​ I think you need to expand on the explanation of the problem. Are you only writing one row per table? If that's the case you can try using BulkAttributeRemover NullValueMapper to drop any null attributes and then use SchemaScanner to find the new schema of the feature and use that to create your table.

For the Azure table we are only righting one row with say columns "High", "Medium", "Low". But given the data, there may be no values for "Medium" and therefore do not want to write to that column in Azure. I have wrestled with BulkAttributeRemover for a bit (probably user error) but have been unable to dynamically remove say "High" and "Low" without the attribute still saying "<missing>".

use python caller.

if the attribute is null

• fmeobjects.FMEFeature.removeAttribute

then pass the feature to py output

The code i currently have in my PythonCaller is as follows. The issue that I am having it that I then manually have to select "Attributes to Hide" as that attribute will still be visible with values of <missing> and sent to the Azure Table Storage writer, leading it to enter blank strings into the table.

import fme

import fmeobjects

class FeatureProcessor(object):

def __init__(self):

pass

def input(self, feature):

for tmpAttrName in feature.getAllAttributeNames():

tmpAttrVal = feature.getAttribute(tmpAttrName)

if(tmpAttrVal is None) or (tmpAttrVal == 99):

feature.removeAttribute(tmpAttrName)

self.pyoutput(feature)

def close(self):

pass

@svteshipley​ If an attribute is flagged as <missing> then the SchemaScanner will ignore it.  So if you start with this:

_creation_instance	0
attr1	a
attr2	b
attr3	<null>

and pass it through NullAttributeMapper and set <null> to <missing> then the SchemaScanner will give you this:

attribute{0}.fme_data_type	fme_varchar(2)
attribute{0}.name	attr1
attribute{1}.fme_data_type	fme_varchar(2)
attribute{1}.name	attr2
attribute{2}.fme_data_type	fme_uint8
attribute{2}.name	_creation_instance

Note the absence  of attr3 in the schema list

_creation_instance	0
attr1	a
attr2	b
attr3	<null>

attribute{0}.fme_data_type	fme_varchar(2)
attribute{0}.name	attr1
attribute{1}.fme_data_type	fme_varchar(2)
attribute{1}.name	attr2
attribute{2}.fme_data_type	fme_uint8
attribute{2}.name	_creation_instance

The SchemaScanner was exactly what was needed. I had to do some research into it to understand. For future reference, I created an Azure Table Storage writer and selected dynamic. From the SchemaScanner I connected the <Schema> and Output to the writer. Within the writer I manually entered the table name to write to, had Schema Sources as "Schema From Schema Feature", Schema Definition Name as "fme_feature_type_name" and for Attributes to Remove as "azure_table_partitionkey, and azure_table_rowkey". These attributes were selected to remove as I had them setup custom and otherwise they would also be written to the table. Now in the User Attributes tab it had the default row of type text.