Discussion: Removing "baseline" branches from names

[jhclark]

Consider this task from the cdec_kftt.tape example:

task align_dir : cdec
                < src=(AlignDir: s2t=$f_train@download t2s=$e_train@download)
                < tgt=(AlignDir: s2t=$e_train@download t2s=$f_train@download)
                > align=align.directional.gz {                                                                                                  
  paste $src $tgt | sed 's/\t/ ||| /g' > corpus                                                                                                      
  $cdec/training/model1 -A -d -v corpus | gzip > align.directional.gz
}

While there are 2 realizations here, which will be used by symmetrization via grafting, the directory names will be a bit unintuitive:

• Baseline.baseline
• AlignDir.t2s

This is because the first branch of each branch point is considered the baseline. And baseline branches are omitted from directory names for 2 reasons:

• To keep realization names from getting crazy long
• But more importantly, so that when new branch points are added, a new element isn't suddenly added to the realization directory name. This way, it's easy to add branch points to existing workflows.

However, this isn't terribly intuitive. This should be well-documented.

Perhaps there's another way of making this clearer to the user as well. Thoughts?

[dowobeha]

@jhclark, does the resolution of #96 also resolve this issue?

[jhclark]

Yes it does.