[ADAM-1480] Add switch to disable the fast concat method.

[fnothaft]

Resolves #1478.

[coveralls]

Coverage increased (+0.08%) to 81.716% when pulling 9f4c33b on fnothaft:issues/1478-concat-fallback into 93b32c6 on bigdatagenomics:master.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/ADAM-prb/1936/
Test PASSed.

[heuermh]

Not a fan of double-negatives; is there a better way to do args4j flags that default to true?

[heuermh]

Per this, #1295, and #1438, perhaps we should brainstorm on how to clean up the save methods and ensure consistency. One or more enums, better SaveArgs classes, a context map, etc.

[heuermh]

I'd be willing to merge this without changes. It adds more pressure towards a future refactoring of save methods for consistency though.

[fnothaft]

@heuermh with #1487 in the pipe, shall we merge this as is?

[fnothaft]

Ping.

[coveralls]

Coverage decreased (-0.2%) to 81.787% when pulling fe0b50b on fnothaft:issues/1478-concat-fallback into 93d17b4 on bigdatagenomics:master.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/ADAM-prb/1957/
Test PASSed.

[heuermh]

Not a fan of double-negatives; is there a better way to do args4j flags that default to true?

What I'm thinking is something like this
http://stackoverflow.com/questions/15008758/parsing-boolean-values-with-argparse/31347222#31347222

where instead of -disable arguments that default to false, we have positive and negative arguments in a mutally exclusive group with the positive one being defaulted to true. That way the field can be named in the positive and code reads better. I'm not quite sure how to set up such with Args4j though.

[fnothaft]

I don't really see the -disable "negative" name as an issue, and I don't have the bandwidth to track down how to do what you're talking about with Args4j. Can we merge this as is?

[heuermh]

Args4j can simulate mutually exclusive groups with forbids, however I don't see a way of combining two --feature and --no-feature style options into a single value as I describe above.

An alternative would be to use ExplicitBooleanOptionHandler and default to true, i.e.

@Args4jOption
  (
    required = false,
    name = "-fast_concat",
    usage = "Enable the parallel file concatenation engine. Defaults to true.",
    handler = classOf[ExplicitBooleanOptionHandler]
  )
var fastConcat: Boolean = true

Then these would be true

$ adam-submit -- adam2vcf -single foo.adam foo.vcf
$ adam-submit -- adam2vcf -single -fast_concat true foo.adam foo.vcf

and this would be false

$ adam-submit -- adam2vcf -single -fast_concat false foo.adam foo.vcf

and this would fail

$ adam-submit -- adam2vcf -single -fast_concat foo.adam foo.vcf
"foo.adam" is not a legal boolean value
 VCF    : The VCF file to convert
 ADAM   : Location to write ADAM Variant data
...

Would that be an improvement?

[heuermh]

Created #1503 for further discussion

[heuermh]

Thank you, @fnothaft!