Add format_map to pipeline_tools code (#171)

* Add format_map to pipeline_tools code

* Fix getting format map

pipeline_tools/shared/submission/create_analysis_metadata.py

@@ -8,6 +8,7 @@
 from google.cloud import storage
 import re
 import arrow
+from pipeline_tools.shared.submission.format_map import EXTENSION_TO_FORMAT
 
 
 def create_analysis_process(
@@ -511,20 +512,11 @@ def main():
         required=True,
         help='Path to JSON file containing info about outputs.',
     )
-    parser.add_argument(
-        '--format_map',
-        required=True,
-        help='JSON file providing map of file extensions to formats.',
-    )
     parser.add_argument(
         '--add_md5s', help='Set to "true" to add md5 checksums to file metadata'
     )
     args = parser.parse_args()
 
-    # Get the extension_to_format mapping
-    with open(args.format_map) as f:
-        extension_to_format = json.load(f)
-
     schema_url = args.schema_url.strip('/')
 
     # Get metadata for inputs and outputs
@@ -533,7 +525,7 @@ def main():
         output_urls = f.read().splitlines()
     outputs = get_outputs(
         output_urls=output_urls,
-        extension_to_format=extension_to_format,
+        extension_to_format=EXTENSION_TO_FORMAT,
         schema_url=schema_url,
         analysis_file_version=args.analysis_file_version,
     )

pipeline_tools/shared/submission/format_map.py

@@ -0,0 +1,21 @@
+EXTENSION_TO_FORMAT = {
+    "[.]bam$": "bam",
+    "[_]metrics$": "metrics",
+    "[.]txt$": "txt",
+    "[.]log$": "log",
+    "[.]pdf$": "pdf",
+    "[.]results$": "results",
+    "[.]theta$": "theta",
+    "[.]cnt$": "cnt",
+    "[.]time$": "time",
+    "[.]model$": "model",
+    "[.]bai$": "bai",
+    "[.]tsv$": "tsv",
+    "[.]mtx$": "mtx",
+    "[.]h5$": "h5",
+    "[.]csv$": "csv",
+    "[.]csv.gz$": "csv.gz",
+    "[.]zarr": "matrix",
+    "[.]npz$": "npz",
+    "[.]npy$": "npy",
+}