Revise 'input_check' to 'input_assure'; enforce JSON key alteration to match the sample ID if a mismatch is detected

bin/input_assure.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python
+
+import json
+import argparse
+import csv
+import gzip
+import sys
+
+
+def open_file(file_path, mode):
+    # Open a file based on the file extension
+    if file_path.endswith(".gz"):
+        return gzip.open(file_path, mode)
+    else:
+        return open(file_path, mode)
+
+
+def check_inputs(json_file, sample_id, address, output_error_file, output_json_file):
+    with open_file(json_file, "rt") as f:
+        json_data = json.load(f)
+
+    # Define a variable to store the match_status (True or False)
+    match_status = sample_id in json_data
+
+    # Initialize the error message
+    error_message = None
+
+    # Check for multiple keys in the JSON file and define error message
+    keys = list(json_data.keys())
+    original_key = keys[0] if keys else None
+
+    if len(keys) == 0:
+        error_message = f"{json_file} is completely empty!"
+        print(error_message)
+        sys.exit(1)
+    elif len(keys) > 1:
+        # Check if sample_id matches any key
+        if not match_status:
+            error_message = f"No key in the MLST JSON file ({json_file}) matches the specified sample ID '{sample_id}'. The first key '{original_key}' has been forcefully changed to '{sample_id}' and all other keys have been removed."
+            # Retain only the specified sample ID
+            json_data = {sample_id: json_data.pop(original_key)}
+        else:
+            error_message = f"MLST JSON file ({json_file}) contains multiple keys: {keys}. The MLST JSON file has been modified to retain only the '{sample_id}' entry"
+            # Remove all keys expect the one matching sample_id
+            json_data = {sample_id: json_data[sample_id]}
+    elif not match_status:
+        # Define error message based on meta.address (query or reference)
+        if address == "null":
+            error_message = f"Query {sample_id} ID and JSON key in {json_file} DO NOT MATCH. The '{original_key}' key in {json_file} has been forcefully changed to '{sample_id}': User should manually check input files to ensure correctness."
+        else:
+            error_message = f"Reference {sample_id} ID and JSON key in {json_file} DO NOT MATCH. The '{original_key}' key in {json_file} has been forcefully changed to '{sample_id}': User should manually check input files to ensure correctness."
+        # Update the JSON file with the new sample ID
+        json_data[sample_id] = json_data.pop(original_key)
+
+    # Write file containing relevant error messages
+    if error_message:
+        with open(output_error_file, "w", newline="") as f:
+            writer = csv.writer(f)
+            writer.writerow(["sample", "JSON_key", "error_message"])
+            writer.writerow([sample_id, keys, error_message])
+
+    # Write the updated JSON data back to the original file
+    with gzip.open(output_json_file, "wt") as f:
+        json.dump(json_data, f, indent=4)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Check sample inputs, force change if ID ≠ KEY, and generate an error report."
+    )
+    parser.add_argument("--input", help="Path to the mlst.json file.", required=True)
+    parser.add_argument(
+        "--sample_id", help="Sample ID to check in the JSON file.", required=True
+    )
+    parser.add_argument(
+        "--address", help="Address to use in the error message.", required=True
+    )
+    parser.add_argument(
+        "--output_error", help="Path to the error report file.", required=True
+    )
+    parser.add_argument(
+        "--output_json", help="Path to the MLST JSON file (gzipped).", required=True
+    )
+
+    args = parser.parse_args()
+
+    check_inputs(
+        args.input, args.sample_id, args.address, args.output_error, args.output_json
+    )

conf/modules.config

@@ -13,8 +13,6 @@
 process {
 
     // Publish directory names
-    assembly_directory_name = "assembly"
-    summary_directory_name = "summary"
     profile_dists_directory_name = "distances"
     gas_call_directory_name = "call"
 
@@ -27,6 +25,10 @@ process {
         saveAs: { filename -> filename.equals('versions.yml') ? null : filename }
     ]
 
+    withName: INPUT_ASSURE {
+        fair = true
+    }
+
     withName: LOCIDEX_MERGE_REF {
         publishDir = [
             path: locidex_merge_ref_directory_name,

modules/local/input_check/main.nf → modules/local/input_assure/main.nf

@@ -1,7 +1,6 @@
-process INPUT_CHECK{
-    tag "Check Sample Inputs and Generate Error Report"
+process INPUT_ASSURE {
+    tag "Assures Inputs are Consistent"
     label 'process_single'
-    fair true
 
     container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?
         'https://depot.galaxyproject.org/singularity/python:3.8.3' :
@@ -11,19 +10,19 @@ process INPUT_CHECK{
     tuple val(meta), path(mlst)
 
     output:
-    tuple val(meta), path("${meta.id}_match.txt"), path(mlst),      emit: match
+    tuple val(meta), path("${meta.id}.mlst.json.gz"),               emit: result
     tuple val(meta), path("*_error_report.csv"), optional: true,    emit: error_report
     path("versions.yml"),                                           emit: versions
 
     script:
 
     """
-    input_check.py \\
+    input_assure.py \\
         --input ${mlst} \\
         --sample_id ${meta.id} \\
         --address ${meta.address} \\
         --output_error ${meta.id}_error_report.csv \\
-        --output_match ${meta.id}_match.txt
+        --output_json ${meta.id}.mlst.json.gz
 
     cat <<-END_VERSIONS > versions.yml
     "${task.process}":

tests/data/irida/mismatched_iridanext.output.json

@@ -0,0 +1,27 @@
+{
+    "files": {
+        "global": [],
+        "samples": {
+            "sampleR": [
+                {
+                    "path": "input/sampleR_error_report.csv"
+                }
+            ],
+            "sample2": [
+                {
+                    "path": "input/sample2_error_report.csv"
+                }
+            ]
+        }
+    },
+    "metadata": {
+        "samples": {
+            "sampleQ": {
+                "address": "2.2.3"
+            },
+            "sampleR": {
+                "address": "2.2.3"
+            }
+        }
+    }
+}

tests/data/irida/test2_iridanext.output.json → tests/data/irida/multiplekeys_iridanext.output.json

@@ -2,9 +2,9 @@
     "files": {
         "global": [],
         "samples": {
-            "sampleR": [
+            "sample3": [
                 {
-                    "path": "input/sampleR_error_report.csv"
+                    "path": "input/sample3_error_report.csv"
                 }
             ]
         }

tests/data/reports/sample2_empty.mlst.json

@@ -0,0 +1 @@
+{}

tests/data/reports/sample3_multiplekeys.mlst.json

@@ -0,0 +1,12 @@
+{
+    "extra_key": {
+        "l1": "1",
+        "l2": "1",
+        "l3": "2"
+    },
+    "sample3": {
+        "l1": "1",
+        "l2": "1",
+        "l3": "2"
+    }
+}

tests/data/reports/sample3_multiplekeys_nomatch.mlst.json

@@ -0,0 +1,12 @@
+{
+    "sample4": {
+        "l1": "1",
+        "l2": "1",
+        "l3": "2"
+    },
+    "extra_key": {
+        "l1": "1",
+        "l2": "1",
+        "l3": "2"
+    }
+}

tests/data/samplesheets/samplesheet_test2.csv → tests/data/samplesheets/samplesheet-mismatched_IDs.csv

@@ -1,7 +1,7 @@
 sample,mlst_alleles,address
-sampleR,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleF.mlst.json,
 sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleQ.mlst.json,
+sampleR,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleF.mlst.json,
 sample1,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample1.mlst.json,1.1.1
-sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample2.mlst.json,1.1.1
+sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample7.mlst.json,1.1.1
 sample3,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample3.mlst.json,1.1.2
 

tests/data/samplesheets/samplesheet-multiple_keys.csv

@@ -0,0 +1,5 @@
+sample,mlst_alleles,address
+sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleQ.mlst.json,
+sample1,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample1.mlst.json,1.1.1
+sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample2.mlst.json,1.1.1
+sample3,https://raw.githubusercontent.com/phac-nml/gasnomenclature/input_assure/tests/data/reports/sample3_multiplekeys.mlst.json,1.1.2

tests/data/samplesheets/samplesheet-multiplekeys_nomatch.csv

@@ -0,0 +1,5 @@
+sample,mlst_alleles,address
+sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleQ.mlst.json,
+sample1,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample1.mlst.json,1.1.1
+sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample2.mlst.json,1.1.1
+sample3,https://raw.githubusercontent.com/phac-nml/gasnomenclature/input_assure/tests/data/reports/sample3_multiplekeys_nomatch.mlst.json,1.1.2

tests/data/samplesheets/samplesheet_test1.csv → tests/data/samplesheets/samplesheet_emptyJSON.csv

@@ -1,5 +1,6 @@
 sample,mlst_alleles,address
-sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleF.mlst.json,
+sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleQ.mlst.json,
 sample1,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample1.mlst.json,1.1.1
-sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample7.mlst.json,1.1.1
+sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/input_assure/tests/data/reports/sample2_empty.mlst.json,1.1.1
 sample3,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample3.mlst.json,1.1.2
+

tests/data/samplesheets/samplesheet_gzip.csv

@@ -0,0 +1,5 @@
+sample,mlst_alleles,address
+sampleQ,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sampleQ.mlst.json,
+sample1,https://raw.githubusercontent.com/phac-nml/gasnomenclature/input_assure/tests/data/reports/sample1.mlst.json.gz,1.1.1
+sample2,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample2.mlst.json,1.1.1
+sample3,https://raw.githubusercontent.com/phac-nml/gasnomenclature/dev/tests/data/reports/sample3.mlst.json,1.1.2