Merge list of tools to keep and to exclude in 1 file and extract ts f…

…iltered tools

README.md

@@ -95,17 +95,16 @@ The script will generate a TSV file with each tool found in the list of GitHub r
 
 1. Run the extraction as explained before
 2. (Optional) Create a text file with ToolShed categories for which tools need to be extracted: 1 ToolShed category per row ([example for microbial data analysis](data/microgalaxy/categories))
-3. (Optional) Create a text file with list of tools to exclude: 1 tool id per row ([example for microbial data analysis](data/microgalaxy/tools_to_exclude))
-4. (Optional) Create a text file with list of tools to really keep (already reviewed): 1 tool id per row ([example for microbial data analysis](data/microgalaxy/tools_to_keep))
+3. (Optional) Create a TSV (tabular) file with 2 columns: ToolShed ids of tool suites (one per line), Boolean with True to keep and False to exclude / 1 tool id per row ([example for microbial data analysis](data/microgalaxy/tools_to_keep_exclude.tsv))
 4. Run the tool extractor script
 
     ```
-    $ python bin/extract_galaxy_tools.py \
+    $ python bin/extract_galaxy_tools.py filtertools \
         --tools <Path to CSV file with all extracted tools> \
-        --filtered_tools <Path to output CSV file with filtered tools> \
+        --ts_filtered_tools <Path to output TSV with tools filtered based on ToolShed category>
+        --filtered_tools <Path to output TSV with filtered tools based on ToolShed category and manual curation> \
         [--categories <Path to ToolShed category file>] \
-        [--excluded <Path to excluded tool file category file>]\
-        [--keep <Path to to-keep tool file category file>]
+        [--keep_exclude <Path to a TSV file with 2 columns: ToolShed ids of tool suites (one per line), Boolean with True to keep and False to exclude>]
     ```
 
 ## Development

bin/extract_galaxy_tools.py

@@ -559,12 +559,14 @@ def export_tools(
     :param output_fp: path to output file
     :param format_list_col: boolean indicating if list columns should be formatting
     """
-    df = pd.DataFrame(tools)
+    df = (
+        pd.DataFrame(tools)
+        .sort_values("Galaxy wrapper id")
+    )
     if format_list_col:
         df["ToolShed categories"] = format_list_column(df["ToolShed categories"])
         df["EDAM operation"] = format_list_column(df["EDAM operation"])
         df["EDAM topic"] = format_list_column(df["EDAM topic"])
-
         df["bio.tool ids"] = format_list_column(df["bio.tool ids"])
 
         # the Galaxy tools need to be formatted for the add_instances_to_table to work
@@ -580,30 +582,31 @@ def export_tools(
 def filter_tools(
     tools: List[Dict],
     ts_cat: List[str],
-    excluded_tools: List[str],
-    keep_tools: List[str],
-) -> List[Dict]:
+    keep_excl_tools: List[Dict],
+) -> tuple:
     """
     Filter tools for specific ToolShed categories and add information if to keep or to exclude
 
     :param tools: dictionary with tools and their metadata
     :param ts_cat: list of ToolShed categories to keep in the extraction
     :param excluded_tools: list of tools to skip
-    :param keep_tools: list of tools to keep
+    :param keep_excl_tools: dictionary with tools and their status (True to keep, False to exclude)
     """
+    ts_filtered_tools = []
     filtered_tools = []
     for tool in tools:
         # filter ToolShed categories and leave function if not in expected categories
         if check_categories(tool["ToolShed categories"], ts_cat):
             name = tool["Galaxy wrapper id"]
-            tool["Reviewed"] = name in keep_tools or name in excluded_tools
-            tool["To keep"] = None
-            if name in keep_tools:
-                tool["To keep"] = True
-            elif name in excluded_tools:
-                tool["To keep"] = False
-            filtered_tools.append(tool)
-    return filtered_tools
+            tool["Reviewed"] = name in keep_excl_tools
+            keep_status = None
+            if name in keep_excl_tools:
+                keep_status = keep_excl_tools[name][1]
+                if keep_status:
+                    filtered_tools.append(tool)
+            tool["To keep"] = keep_status
+            ts_filtered_tools.append(tool)
+    return ts_filtered_tools, filtered_tools
 
 
 if __name__ == "__main__":
@@ -632,30 +635,31 @@ def filter_tools(
     filtertools = subparser.add_parser("filtertools", help="Filter tools")
     filtertools.add_argument(
         "--tools",
-        "-t",
+        "-i",
         required=True,
         help="Filepath to TSV with all extracted tools, generated by extractools command",
     )
+    filtertools.add_argument(
+        "--ts_filtered_tools",
+        "-t",
+        required=True,
+        help="Filepath to TSV with tools filtered based on ToolShed category",
+    )
     filtertools.add_argument(
         "--filtered_tools",
         "-f",
         required=True,
-        help="Filepath to TSV with filtered tools",
+        help="Filepath to TSV with tools filtered based on ToolShed category and manual curation",
     )
     filtertools.add_argument(
         "--categories",
         "-c",
         help="Path to a file with ToolShed category to keep in the extraction (one per line)",
     )
     filtertools.add_argument(
-        "--exclude",
-        "-e",
-        help="Path to a file with ToolShed ids of tools to exclude (one per line)",
-    )
-    filtertools.add_argument(
-        "--keep",
+        "--keep_exclude",
         "-k",
-        help="Path to a file with ToolShed ids of tools to keep (one per line)",
+        help="Path to a TSV file with 2 columns: ToolShed ids of tool suites (one per line), Boolean with True to keep and False to exclude ",
     )
     args = parser.parse_args()
 
@@ -681,11 +685,11 @@ def filter_tools(
         export_tools(tools, args.all_tools, format_list_col=True, add_usage_stats=True)
 
     elif args.command == "filtertools":
-        tools = pd.read_csv(Path(args.tools), sep="\t", keep_default_na=False).to_dict("records")
+        tools = pd.read_csv(args.tools, sep="\t", keep_default_na=False).to_dict("records")
         # get categories and tools to exclude
         categories = read_file(args.categories)
-        excl_tools = read_file(args.exclude)
-        keep_tools = read_file(args.keep)
+        keep_excl_tools = pd.read_csv(args.keep_exclude, sep="\t", index_col=0, header=None).to_dict("index")
         # filter tool lists
-        filtered_tools = filter_tools(tools, categories, excl_tools, keep_tools)
+        ts_filtered_tools, filtered_tools = filter_tools(tools, categories, keep_excl_tools)
+        export_tools(ts_filtered_tools, args.ts_filtered_tools)
         export_tools(filtered_tools, args.filtered_tools)

bin/get_community_tools.sh

@@ -12,10 +12,10 @@ for com_data_fp in data/communities/* ; do
                 python bin/extract_galaxy_tools.py \
                         filtertools \
                         --tools "results/all_tools.tsv" \
+                        --ts_filtered_tools "results/$community/tools_filtered_by_ts_categories.tsv" \
                         --filtered_tools "results/$community/tools.tsv" \
                         --categories "data/communities/$community/categories" \
-                        --exclude "data/communities/$community/tools_to_exclude" \
-                        --keep "data/communities/$community/tools_to_keep"
+                        --keep_exclude "data/communities/$community/tools_to_keep_exclude.tsv"
 
                 python bin/create_interactive_table.py \
                         --table "results/$community/tools.tsv" \