added predicate option to loadCoverage

[akmorrow13]

No description provided.

[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/1474/
Test PASSed.

[fnothaft]

Ah, so this is a philosophical decision articulated many moons in the past by @laserson, but we don't have the predicates on the "automagical" functions (i.e., loadX, where X is not a specific file format). IIRC, the reasoning here is that the way the predicate is executed is hard to reason about without knowing the underlying file format, so thus it isn't good practice to "move" this underneath the user. We used to have another function that would apply a predicate to raw Avro, but we don't have this now, which also means that the records loaded from this function would be different, since the predicate would be applied to a Parquet file, but not to coverage loaded from (e.g.) BED.

What's the use case here? I know you had one in mind, so I'm wondering what's the best way to move this forward.

[Georgehe4]

The use case here is to enable loading in custom coverage files in Mango, since users might only be interested in viewing coverage information.

[fnothaft]

The use case here is to enable loading in custom coverage files in Mango, since users might only be interested in viewing coverage information.

Those are just stored as Parquet Feature files, right? If so, then I'd do loadParquetFeatures with the predicate. I'm guessing that the goal is to have a region based predicate? If so, perhaps we should refactor the signature of the changed method (to take a region to filter on), and then add a filterByOverlappingReferenceRegion in the non-Parquet fork of the code (https://github.com/bigdatagenomics/adam/pull/1156/files#diff-d36ea7d0742decd0b040a73a96af06e9R972). How does that sound?

[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/1482/
Test PASSed.

[akmorrow13]

I added a loadParquetCoverage and loadCoverage, the first which accepts a predicate. This is how loadFeatures is implemented.

[fnothaft]

Thanks for updating this @akmorrow13! This LGTM. I will leave open for a day for any other review comments and will merge if there are no objections.

[fnothaft]

Thanks @akmorrow13! I've merged this as 5e2853e.