Null Attributes

Null attributes are a relatively new, but very important, part of FME’s attribute handling.

What is a Null Value?

In general, a null attribute value is the equivalent of nothing. However, it’s important to be precise in our terminology when we talk about “null”. In reality there are many ways to represent nothing:

• An attribute is null
• An attribute exists but is empty
• An attribute doesn’t exist (i.e. is missing)
• An attribute is NaN (Not a Number)
• A numeric attribute has a value of zero

In fact, Safe Software’s developers have identified fifteen (15) different ways for “nothing” to be represented in spatial and tabular data.

So when we talk about null in FME, it has a particular meaning. For us, a null is an actual value that is deliberately set to signify that the information does not exist. It tells us that the lack of information is not a mistake – as a missing or empty value might.

How does FME Represent Null Values? FME’s internal engine has its own state to represent null. However, when presented to the user, a null value is usually represented as .

For example, this feature in the Logger has for a number of attributes:

Similarly, the FME Data Inspector will depict nulls as :

It can also differentiate between states by displaying or as well.

Recognizing Null Values

When FME reads data, if the source attributes contain nulls – and the Reader format has been updated to support them – then FME will emit that attribute with a null value.

To check for incoming nulls the Tester transformer has a specific operator to test for null, empty, and missing values:

Because the Tester interface is incorporated into many facets of FME (such as the TestFilter transformer) you can test for nulls wherever you find that interface.

Other transformers, such as the Matcher, also allow testing for nulls. In the case of the Matcher there is even a parameter that can be used to decide whether null, empty, and missing values should be treated equally, or as their own unique representation:

Setting a Null Value

The usual way to set an attribute value is with the AttributeCreator, and this has an option in its drop-down menu to set a value to null:

When you set an attribute to null, and send it to a Writer, then what happens depends upon the data format.

If the format supports nulls – and the Writer has been updated to support them too – then the destination dataset will contain null attribute values.

If the format doesn’t support nulls, then FME will automatically convert the data to the closest representation that is supported.

Bulk Null Updates

The way to handle bulk updates of attributes is with the NullAttributeMapper transformer.

The NullAttributeMapper transformer allows the author to check values for any or all attributes on a feature, and convert them in bulk to or from null.

For example, here the author is checking for all attributes that are either missing or empty, and converting them to nulls:

Here the author is checking a specific attribute for existing null values. If the value is set to null then it gets replaced with a zero. Presumably this must be a numeric field. If it was a text field perhaps instead the author would set it to an empty string:

In this workspace a colleague is trying to write out a list of parks to a Geodatabase dataset. It’s important to them that the parks are in alphabetical order – according to their name – and that features with no park names are written as null and appear last in the dataset.

However, the workspace they have does not seem to be doing what they need. The parks are sorted alphabetically, but un-named parks always appear first.

1) Start Workbench

Open the workspace C:\FMEData2015\Workspaces\DesktopAdvanced\Exercise5d-Begin.fmw

Inspect the source dataset by right-clicking the source feature type and choosing Inspect.

In the Data Inspector examine the data in the Table View window. You’ll see that the data is in order of ID, not name and that there are values scattered throughout.

2) Add NullAttributeMapper

Add a NullAttributeMapper transformer prior to the Sorter transformer.

Open the parameters dialog.

Ensure “Map” is set to Selected Attributes, and choose the attribute ParkName.

Underneath that is a section of what to map to.

We know the values in here are currently listed as so set the “If Attribute Value Is” parameter to Missing (Selected Attributes Only)

We want to map these to a value that will appear at the bottom of any alphabetically sorted list, so change “Map To” to New Value and enter ZZZZ as the new value. Click OK to close the dialog.

3) Add NullAttributeMapper

Now add a second NullAttributeMapper; this time it should be connected after the Sorter.

Open the dialog and, once again, ensure “Map” is set to Selected Attributes and select the ParkName attribute. Then turn the ZZZZ values back to nulls.

Technically we could just turn them back into ; the Geodatabase Writer will write them out as nulls. However, assuming we didn’t know that, null is the safer option and bound to give us what we want.

4) Save and Run Workspace

Save the workspace and then run it. Inspect the output. This time the data should be sorted by ParkName, but with all null values at the end of the dataset: