C2W2: How to get number of examples from artifacts?

Hey, I believe I’m missing out on something conceptual about data handling. As I’ve learned, ExampleGen is supposed to get data in our system and produce artefacts along the way. Artefacts have a bunch of attributes, such as .span or .uri. But is it possible to get something useful out of them (aka the number of events in train/eval, for example) having the artefact and the ExampleGen objects at hand? And if not, why? Is the developer expected to estimate the numbers always from the raw dataset? Deeper in the pipeline, one can add further splitting that would be a bit error-prone to just carry in mind, so I believe there is something more intelligent implemented. Can anyone advise?

Please see apache beam to understand the way ingestion works in big data world.
Components talk to each other via the metadata store. ExampleGen acts as the input to the downstream compoents.

Use StatisticsGen for generating summary statistics about the dataset.
The ungraded labs and assignment cover this in good detail. Give it a shot.