From 2ec2e386bc2cf132e547689633bd498792ee204b Mon Sep 17 00:00:00 2001
From: William Rowell <wrowell@pacificbiosciences.com>
Date: Fri, 19 Jul 2024 21:22:02 -0700
Subject: [PATCH] Added tests to compare md5sums for VCF or SAM records while
 ignoring variable headers.

---
 .../wdl_tests/calculate_sam_record_md5sum.wdl | 49 +++++++++++++++++++
 .../wdl_tests/calculate_vcf_record_md5sum.wdl | 49 +++++++++++++++++++
 2 files changed, 98 insertions(+)
 create mode 100644 src/wdlci/wdl_tests/calculate_sam_record_md5sum.wdl
 create mode 100644 src/wdlci/wdl_tests/calculate_vcf_record_md5sum.wdl

diff --git a/src/wdlci/wdl_tests/calculate_sam_record_md5sum.wdl b/src/wdlci/wdl_tests/calculate_sam_record_md5sum.wdl
new file mode 100644
index 0000000..ac83833
--- /dev/null
+++ b/src/wdlci/wdl_tests/calculate_sam_record_md5sum.wdl
@@ -0,0 +1,49 @@
+version 1.0
+
+# Compare md5sums of SAM records in input files, e.g., BAMs or CRAMs while ignoring headers
+# Input type: SAM/BAM/CRAM
+
+task calculate_sam_record_md5sum {
+	input {
+		File current_run_output
+		File validated_output
+	}
+
+	Int disk_size = ceil(size(current_run_output, "GB") + size(validated_output, "GB") + 50)
+
+	command <<<
+		set -euo pipefail
+
+		err() {
+			message=$1
+
+			echo -e "[ERROR] $message" >&2
+		}
+
+		# Compare files
+		echo "Comparing SAM record md5sums"
+		current_run_md5sum=$(samtools view ~{current_run_output} | md5sum | cut -d ' ' -f 1)
+		validated_output_md5sum=$(samtools view ~{validated_output} | md5sum | cut -d ' ' -f 1)
+
+		if [[ "$current_run_md5sum" != "$validated_output_md5sum" ]]; then
+			err "SAM record md5sums did not match:
+				Expected md5sum: [$validated_output_md5sum]
+				Current run md5sum: [$current_run_md5sum]"
+			exit 1
+		else
+			echo "SAM record md5sums matched for file [~{basename(validated_output)}]"
+		fi
+	>>>
+
+	output {
+	}
+
+	runtime {
+		docker: "dnastack/dnastack-wdl-ci-tools:0.0.1"
+		cpu: 1
+		memory: "3.75 GB"
+		disk: disk_size + " GB"
+		disks: "local-disk " + disk_size + " HDD"
+		preemptible: 1
+	}
+}
diff --git a/src/wdlci/wdl_tests/calculate_vcf_record_md5sum.wdl b/src/wdlci/wdl_tests/calculate_vcf_record_md5sum.wdl
new file mode 100644
index 0000000..071aed0
--- /dev/null
+++ b/src/wdlci/wdl_tests/calculate_vcf_record_md5sum.wdl
@@ -0,0 +1,49 @@
+version 1.0
+
+# Compare md5sums of VCF records in input files, ignoring headers
+# Input type: VCF/BCF/VCF.gz/BCF.gz
+
+task calculate_vcf_record_md5sum {
+	input {
+		File current_run_output
+		File validated_output
+	}
+
+	Int disk_size = ceil(size(current_run_output, "GB") + size(validated_output, "GB") + 50)
+
+	command <<<
+		set -euo pipefail
+
+		err() {
+			message=$1
+
+			echo -e "[ERROR] $message" >&2
+		}
+
+		# Compare files
+		echo "Comparing VCF record md5sums"
+		current_run_md5sum=$(bcftools view -H ~{current_run_output} | md5sum | cut -d ' ' -f 1)
+		validated_output_md5sum=$(bcftools view -H ~{validated_output} | md5sum | cut -d ' ' -f 1)
+
+		if [[ "$current_run_md5sum" != "$validated_output_md5sum" ]]; then
+			err "VCF record md5sums did not match:
+				Expected md5sum: [$validated_output_md5sum]
+				Current run md5sum: [$current_run_md5sum]"
+			exit 1
+		else
+			echo "VCF record md5sums matched for file [~{basename(validated_output)}]"
+		fi
+	>>>
+
+	output {
+	}
+
+	runtime {
+		docker: "dnastack/dnastack-wdl-ci-tools:0.0.1"
+		cpu: 1
+		memory: "3.75 GB"
+		disk: disk_size + " GB"
+		disks: "local-disk " + disk_size + " HDD"
+		preemptible: 1
+	}
+}