fix docker image parsing and purls (#413)

Author: SUSTAPLE117

models/purl.go

@@ -1,6 +1,7 @@
 package models
 
 import (
+	"errors"
 	"fmt"
 	"strings"
 
@@ -68,9 +69,56 @@ func (p *Purl) Link() string {
 	return ""
 }
 
+// PurlFromDockerImage parses a Docker image reference and returns a valid
+// Docker PURL per https://github.com/package-url/purl-spec/blob/main/types/docker-definition.json
+//
+// Examples:
+//
+//	alpine:latest           -> pkg:docker/alpine@latest
+//	ghcr.io/org/image:tag   -> pkg:docker/org/image@tag?repository_url=ghcr.io
+//	myimage@sha256:abcdef   -> pkg:docker/myimage@sha256%3Aabcdef
 func PurlFromDockerImage(image string) (Purl, error) {
-	purl, err := packageurl.FromString("pkg:docker/" + image)
-	return Purl{PackageURL: purl}, err
+	if image == "" {
+		return Purl{}, errors.New("empty docker image reference")
+	}
+
+	var name, version string
+	var qualifiers packageurl.Qualifiers
+
+	// Split off version: either @digest or :tag
+	if idx := strings.Index(image, "@"); idx != -1 {
+		version = image[idx+1:]
+		image = image[:idx]
+	} else if idx := strings.LastIndex(image, ":"); idx != -1 {
+		// Ensure the colon is after the last slash (i.e. it's a tag, not a port/registry part)
+		if slashIdx := strings.LastIndex(image, "/"); idx > slashIdx {
+			version = image[idx+1:]
+			image = image[:idx]
+		}
+	}
+
+	// Split registry from the path.
+	// A registry is present if the first path component contains a dot or colon,
+	// or is "localhost" (standard Docker reference parsing heuristic).
+	parts := strings.SplitN(image, "/", 2)
+	if len(parts) == 2 && (strings.ContainsAny(parts[0], ".:") || parts[0] == "localhost") {
+		registry := parts[0]
+		qualifiers = packageurl.QualifiersFromMap(map[string]string{
+			"repository_url": registry,
+		})
+		image = parts[1]
+	}
+
+	// Split remaining path into namespace and name
+	if idx := strings.LastIndex(image, "/"); idx != -1 {
+		namespace := image[:idx]
+		name = image[idx+1:]
+		p := packageurl.NewPackageURL("docker", namespace, name, version, qualifiers, "")
+		return Purl{PackageURL: *p}, nil
+	}
+
+	p := packageurl.NewPackageURL("docker", "", image, version, qualifiers, "")
+	return Purl{PackageURL: *p}, nil
 }
 
 func PurlFromGithubActions(uses string, sourceGitRepo string, sourceGitRef string) (Purl, error) {

models/purl_test.go

@@ -53,15 +53,15 @@ func TestPurlFromGithubActions(t *testing.T) {
 		},
 		{
 			uses:     "docker://alpine:latest",
-			expected: "pkg:docker/alpine:latest",
+			expected: "pkg:docker/alpine@latest",
 		},
 		{
 			uses:     "docker://ghcr.io/org/owner/image:tag",
-			expected: "pkg:docker/ghcr.io/org/owner/image:tag",
+			expected: "pkg:docker/org/owner/image@tag?repository_url=ghcr.io",
 		},
 		{
 			uses:     "docker://ghcr.io/org/owner/image@sha256:digest",
-			expected: "pkg:docker/ghcr.io/org/owner/image@sha256:digest",
+			expected: "pkg:docker/org/owner/image@sha256:digest?repository_url=ghcr.io",
 		},
 		{
 			uses:  "",

opa/opa_test.go

@@ -43,7 +43,7 @@ func TestOpaBuiltins(t *testing.T) {
 		{
 			builtin:  "purl.parse_docker_image",
 			input:    `"alpine:latest"`,
-			expected: "pkg:docker/alpine:latest",
+			expected: "pkg:docker/alpine@latest",
 		},
 	}
 

opa/rego/poutine/utils.rego

@@ -10,7 +10,6 @@ unpinned_github_action(purl) if {
 
 unpinned_docker(purl) if {
 	startswith(purl, "pkg:docker/")
-	not contains(purl, "@")
 	not regex.match("@sha256:[a-f0-9]{64}", purl)
 }
 

scanner/inventory_scanner_test.go

@@ -81,7 +81,7 @@ func TestRun(t *testing.T) {
 
 	assert.Contains(t, scannedPackage.BuildDependencies, "pkg:githubactions/actions/checkout@v4")
 	assert.Contains(t, scannedPackage.PackageDependencies, "pkg:githubactions/actions/github-script@main")
-	assert.Contains(t, scannedPackage.PackageDependencies, "pkg:docker/alpine:latest")
+	assert.Contains(t, scannedPackage.PackageDependencies, "pkg:docker/alpine@latest")
 	assert.Equal(t, 3, len(scannedPackage.GitlabciConfigs))
 }
 

scanner/inventory_test.go

@@ -27,32 +27,32 @@ func TestPurls(t *testing.T) {
 	require.NoError(t, err)
 
 	purls := []string{
-		"pkg:docker/node:latest",
+		"pkg:docker/node@latest",
 		"pkg:githubactions/hashicorp/vault-action@v3",
 		"pkg:githubactions/actions/checkout@main",
 		"pkg:githubactions/kartverket/github-workflows@main#.github/workflows/run-terraform.yml",
 		"pkg:githubactions/kartverket/github-workflows@v2.2#.github/workflows/run-terraform.yml",
 		"pkg:githubactions/kartverket/github-workflows@v2.7.1#.github/workflows/run-terraform.yml",
-		"pkg:docker/alpine:latest",
+		"pkg:docker/alpine@latest",
 		"pkg:githubactions/actions/github-script@main",
 		"pkg:githubactions/actions/setup-node@v4",
 		"pkg:githubactions/hashicorp/vault-action@v2.1.0",
 		"pkg:githubactions/actions/checkout@v4",
-		"pkg:docker/ruby:3.2",
-		"pkg:docker/postgres:15",
+		"pkg:docker/ruby@3.2",
+		"pkg:docker/postgres@15",
 		"pkg:gitlabci/include/template?file_name=Auto-DevOps.gitlab-ci.yml",
 		"pkg:gitlabci/include/project?file_name=%2Ftemplates%2F.gitlab-ci-template.yml&project=my-group%2Fmy-project&ref=main",
 		"pkg:gitlabci/include/remote?download_url=https%3A%2F%2Fexample.com%2F.gitlab-ci.yml",
 		"pkg:gitlabci/include/component?project=my-org%2Fsecurity-components%2Fsecret-detection&ref=1.0&repository_url=gitlab.example.com",
 		"pkg:githubactions/org/repo@main",
-		"pkg:docker/debian:vuln",
+		"pkg:docker/debian@vuln",
 		"pkg:githubactions/bridgecrewio/checkov-action@main",
 		"pkg:githubactions/org/repo@main#.github/workflows/Reusable.yml",
 		"pkg:azurepipelinestask/DownloadPipelineArtifact@2",
 		"pkg:azurepipelinestask/Cache@2",
 		"pkg:githubactions/org/owner@main#.github/workflows/ci.yml",
 		"pkg:githubactions/actions/checkout@v5",
-		"pkg:docker/node:18",
+		"pkg:docker/node@18",
 		"pkg:githubactions/some/action@v1",
 	}
 	assert.ElementsMatch(t, i.Purls(*scannedPackage), purls)