Sampler

Preserved features are output through this port.

Features that are not preserved are output through this port.

When specified, this parameter affects the behavior of the sampling types. For example, say the Sampler receives a set of 15 features. Using the Group By parameter, the Sampler breaks the features into 3 groups.

• Group 1 has 3 features
• Group 2 has 5 features
• Group 3 has 7 features

This table shows how sampling amount and type affect the Group By results:

Process At End (Blocking): This is the default behavior. Processing will only occur in this transformer once all input is present.

Process When Group Changes (Advanced): This transformer will process input groups in order. Changes of the value of the Group By parameter on the input stream will trigger processing on the currently accumulating group. This may improve overall speed (particularly with multiple, equally-sized groups), but could cause undesired behavior if input groups are not truly ordered.

Considerations for Using Group By

There are two typical reasons for using Process When Group Changes (Advanced) . The first is incoming data that is intended to be processed in groups (and is already so ordered). In this case, the structure dictates Group By usage - not performance considerations.

The second possible reason is potential performance gains.

Performance gains are most likely when the data is already sorted (or read using a SQL ORDER BY statement) since less work is required of FME. If the data needs ordering, it can be sorted in the workspace (though the added processing overhead may negate any gains).

Sorting becomes more difficult according to the number of data streams. Multiple streams of data could be almost impossible to sort into the correct order, since all features matching a Group By value need to arrive before any features (of any feature type or dataset) belonging to the next group. In this case, using Group By with Process At End (Blocking) may be the equivalent and simpler approach.

Note: Multiple feature types and features from multiple datasets will not generally naturally occur in the correct order.

As with many scenarios, testing different approaches in your workspace with your data is the only definitive way to identify performance gains.

Determines the number of features to send to the Sampled port: either a total number of features, or a sampling of features.

If the Sampling Rate (N) is 0, all input data will be sent to the NotSampled port.

For example:

When this parameter is left at the default setting of No, the transformer works with features in the order in which they are received.

If this parameter is Yes, the input features are shuffled before they are sampled. All features are blocked in the Sampler until the last feature arrives. The output ports will return the features in the original order they were received by this transformer.