README.md

Venctester 3 codebase

Directory structure

The codebase consists of the following directories:

• encoders (all encoder-specific code)
• core (all generic code)

Only the encoders directory needs to be touched if support for a new encoder is to be added.

Adding support for a new encoder

To add support for a new encoder, the following things need to be done:

1. Implement a new subclass of tester.encoders.base.EncoderBase
   • This class represents an encoder binary
   • This class should be named <EncoderName>
   • This class should be placed in tester/encoders/<encoder_name>.py
2. Implement a new subclass of tester.encoders.base.EncoderBase.ParamSet
   • This class represents a set of command line parameters to be passed to the encoder when encoding
   • This class should be named <ParamSet>
   • This class should be placed in tester/encoders/<encoder_name>.py
   • This class should be inside the <EnocderName> class

tester/encoders/kvazaar.py and tester/core/test.py together provide a full example.

Adding new metrics and metric calculation

The commit c41a1aa104d219ff783d846b142da80d635bb9bf is a great example on how to both add a new metric and the calculation

• If the metric is an absolute metric the call to the calculation should be added to the _calculate_metrics_for_one_run function in tester.py::Tester
• For comparative metrics the calculation should be added under the SequenceMetrics class in metrics.py and the call should be added into the values_by_field in CsvField.add_entry or the equivalent in table.py if the metric should be included in tables
• The enums in csv.py, table.py, and graph.py need to be updated to include the metric for each output type that should include the metric.
• Also suggested adding default values to the corresponding Cfg() structures.