slow validation caused by string replaceAll to remove nulls

[awgymer]

Fixes #90

It turns out replaceAll to remove null values from the json string representation of the samplesheet is the bottleneck here.

Using JsonGenerator with .excludeNulls to convert the object to a string and back again to a JSONArray solves this.

Using a single-end samplesheet with 60 rows:

before

Oct-18 11:57:31.911 [main] DEBUG nextflow.validation.SchemaValidator - Starting validation: 'samplesheet.se.60.csv' with 'assets/schema_input.json'
Oct-18 11:57:31.916 [main] DEBUG nextflow.validation.SchemaValidator - Removing null
Oct-18 12:07:39.610 [main] DEBUG nextflow.validation.SchemaValidator - nulls removed
Oct-18 12:07:39.613 [main] DEBUG nextflow.validation.SchemaValidator - Validation passed: 'samplesheet.se.60.csv' with 'assets/schema_input.json'

after

Oct-18 12:29:44.364 [main] DEBUG nextflow.validation.SchemaValidator - Starting validation: 'samplesheet.se.60.csv' with 'assets/schema_input.json'
Oct-18 12:29:44.369 [main] DEBUG nextflow.validation.SchemaValidator - Removing null
Oct-18 12:29:44.370 [main] DEBUG nextflow.validation.SchemaValidator - nulls removed
Oct-18 12:29:44.373 [main] DEBUG nextflow.validation.SchemaValidator - Validation passed: 'samplesheet.se.60.csv' with 'assets/schema_input.json'

N.B. Interestingly it should be noted that if you use a yaml input format samplesheet rather than a csv you do not see the same issue. This leads me to suspect that null values in a yaml are being parsed differently at some point before this:

Oct-18 11:47:07.491 [main] DEBUG nextflow.validation.SchemaValidator - Starting validation: 'samplesheet.se.60.yaml' with 'assets/schema_input.json'
Oct-18 11:47:07.502 [main] DEBUG nextflow.validation.SchemaValidator - Removing null
Oct-18 11:47:07.503 [main] DEBUG nextflow.validation.SchemaValidator - nulls removed
Oct-18 11:47:07.503 [main] DEBUG nextflow.validation.SchemaValidator - Validation passed: 'samplesheet.se.60.yaml' with 'assets/schema_input.json'

[nvnieuwk]

Looks good to me! Thanks @awgymer