Echo translation engine doesn't work for Paratext projects

[pmachapman]

This is likely low priority.

The echo translation engine only works with text files. As Scripture Forge now uploads Paratext zip files, we are no longer able to use Echo for testing purposes.

In addition, the language checks 'https://qa.serval-api.org/api/v1/translation/engine-types/echo/languages/en always return isNative: false for echo. Should these always return true, as echo supports any language?

[Enkidu93]

The most straightforward way to accomplish this is to abstract the alignment and inclusion/exclusion of data functionality from NmtPreprocessBuildJob into a separate class and share that with the Echo engine (or have both inherit/implement the same pipeline). This is something that would be more broadly useful, I think. This way the Echo engine could properly reflect the logic in NMT jobs and would be generally more useful for testing. Should we wait on this until the Machine ASP.NET code is moved over to the Serval repo? @ddaspit @johnml1135 What do you guys think? How important is this?

The switch to a default of true is trivial and I think that would make more sense.

[ddaspit]

My plan was that once we move the Machine engine to the Serval repo, we could create a library for shared engine functionality. So, yes, we should wait until we move the Machine engine to the Serval repo.

[Enkidu93]

@johnml1135 Doesn't look like the machine engine has been moved over yet; can we move this back to another release?

[Enkidu93]

Waiting on #451

[Enkidu93]

@ddaspit, do you think it would be appropriate to move some of the logic from PreprocessBuildJob into the service toolkit so that the Echo engine can use it? I'm just not sure how to make that work in a way that maintains a separation between the toolkit and Serval/Machine - or is that degree of separation not a priority? I can just duplicate code, but if we want the Echo engine to be a usable testing engine, we'll need to be conscientious about updating the Echo engine as we make updates to preprocessing particularly. It would be nice if they could just share code so we're always certain that we're testing something as realistic as possible.