@cdxgen/cdxgen 13.2.0 → 13.4.0

package/lib/ecosystems/scalasem.js

@@ -0,0 +1,2618 @@
+import { lstatSync, readFileSync, statSync } from "node:fs";
+import { basename, dirname, isAbsolute, join, resolve, sep } from "node:path";
+import process from "node:process";
+
+import { DEBUG_MODE, readEnvironmentVariable } from "../core/activity.js";
+import { deferFailOnError } from "../core/deferredExit.js";
+import { PROJECT_TYPE_ALIASES } from "../core/env.js";
+import {
+  getAllFiles,
+  getTmpDir,
+  safeCopyFileSync,
+  safeExistsSync,
+  safeMkdirSync,
+  safeMkdtempSync,
+  safeRmSync,
+  safeSpawnSync,
+} from "../core/fs.js";
+import { dirNameStr } from "../core/paths.js";
+import { resolvePluginSourceDir } from "../core/pluginRun.js";
+import { formatDuration } from "../core/processTree.js";
+import { parsetoolsFile } from "../inventory/atomUtils.js";
+import { applyAlgorithmProperties } from "../inventory/cryptoAlgorithmFamily.js";
+import { getJarClasses } from "../inventory/deps.js";
+import {
+  parseScalaArtifact,
+  scalaCoordinateKey,
+  scalaCoordinateOfPurl,
+} from "../inventory/scalaCoords.js";
+import { parsePkgLock } from "./parsers-js.js";
+
+/**
+ * scalasem is the Scala semantic analyzer shipped inside atom-parsetools. This
+ * module is the only place that spawns it, and it owns the projection of the
+ * scalasem report (schema scalasem/2) onto CycloneDX evidence. It follows the
+ * golem and kosi discipline: a failed run may degrade the Scala evidence but
+ * never aborts SBOM generation, and it always says why.
+ */
+
+/** Longest wait for a scalasem run. Large repositories degrade, not hang. */
+const DEFAULT_TIMEOUT_MS = 20 * 60_000;
+
+/** Occurrences kept per component unless CDXGEN_SCALASEM_MAX_OCCURRENCES says otherwise. */
+const DEFAULT_MAX_OCCURRENCES = 200;
+
+// The JDK ships these packages and no component stands for it. `javax.` is
+// split between the JDK and libraries (`javax.servlet`, `javax.inject`), so a
+// `javax.` symbol joins a library only through a class the library ships.
+export const JDK_OWNER_PREFIXES = ["java.", "jdk.", "sun."];
+const JDK_EXTENSION_PREFIX = "javax.";
+
+// The language runtimes every module carries, on each platform. Their
+// occurrences are capped at one per file: they say a file is Scala, not that
+// it uses a library.
+export const SCALA_STDLIB_ARTIFACTS = new Set([
+  "auxlib",
+  "javalib",
+  "nativelib",
+  "scala-library",
+  "scala3-library",
+  "scalajs-library",
+  "scalalib",
+]);
+
+export function appendUniqueProperty(properties, name, value) {
+  if (value === undefined || value === null || value === "") {
+    return;
+  }
+  const propertyValue = String(value);
+  if (
+    !properties.some(
+      (property) => property.name === name && property.value === propertyValue,
+    )
+  ) {
+    properties.push({ name, value: propertyValue });
+  }
+}
+
+export function addPropertyValue(map, key, name, value) {
+  if (!key || value === undefined || value === null || value === "") {
+    return;
+  }
+  map[key] ??= [];
+  appendUniqueProperty(map[key], name, value);
+}
+
+export function addSetValue(map, key, value) {
+  if (!key || !value) {
+    return;
+  }
+  map[key] ??= new Set();
+  map[key].add(value);
+}
+
+/**
+ * True for every project type the Scala path serves, matching the alias list
+ * the BOM generators dispatch on.
+ *
+ * @param {string} language Project language or type.
+ * @returns {boolean}
+ */
+export function isScalasemLanguage(language) {
+  return PROJECT_TYPE_ALIASES.scala.includes(
+    String(language || "").toLowerCase(),
+  );
+}
+
+/**
+ * True when scalasem is turned off, through CDXGEN_SCALASEM_DISABLE or the
+ * --no-scalasem flag, which yargs carries as `scalasem: false`. Library code never reads
+ * process.argv for this.
+ *
+ * @param {Object} options CLI options.
+ * @returns {boolean}
+ */
+export function scalasemDisabled(options = {}) {
+  const raw = (
+    readEnvironmentVariable("CDXGEN_SCALASEM_DISABLE") || ""
+  ).toLowerCase();
+  return (
+    raw === "1" || raw === "true" || raw === "all" || options.scalasem === false
+  );
+}
+
+/**
+ * The scalasem entry point: SCALASEM_CMD wins, then the scalasem.js inside
+ * the installed atom-parsetools package.
+ *
+ * @returns {string|undefined}
+ */
+export function resolveScalasemCommand() {
+  return (
+    readEnvironmentVariable("SCALASEM_CMD") || parsetoolsFile("scalasem.js")
+  );
+}
+
+function scalasemTimeoutMs(options = {}) {
+  const raw =
+    options.scalasemTimeoutMs ??
+    readEnvironmentVariable("CDXGEN_SCALASEM_TIMEOUT");
+  const parsed = Number.parseInt(String(raw ?? ""), 10);
+  return Number.isFinite(parsed) && parsed > 0 ? parsed : DEFAULT_TIMEOUT_MS;
+}
+
+// Time scalasem gets to stop its builds and itself after its own limit,
+// before cdxgen stops it.
+const TIMEOUT_GRACE_MS = 30_000;
+
+// The build files that identify a Scala project, for a source argument that
+// names the working directory relative to itself.
+const SCALA_BUILD_MARKERS = [
+  "build.sbt",
+  "build.sc",
+  "build.mill",
+  "pom.xml",
+  "project.scala",
+];
+
+/**
+ * Where the scalasem report is persisted. An absolute --semantics-slices-file
+ * keeps its directory; anything else lands in the evinse output directory, the
+ * same rule the atom slice writer follows.
+ *
+ * @param {Object} options CLI options carrying `output` and `semanticsSlicesFile`.
+ * @returns {string} Absolute report path.
+ */
+export function scalasemOutputFile(options = {}) {
+  const requested = options.semanticsSlicesFile || "semantics.slices.json";
+  // Comparing with resolve() missed Windows paths written with forward slashes
+  // or without a drive.
+  if (isAbsolute(requested)) {
+    return requested;
+  }
+  let outputDir;
+  if (options.output) {
+    const resolvedOutput = resolve(options.output);
+    let isDir = false;
+    try {
+      isDir =
+        safeExistsSync(resolvedOutput) &&
+        lstatSync(resolvedOutput).isDirectory();
+    } catch (_err) {
+      isDir = false;
+    }
+    outputDir = isDir ? resolvedOutput : dirname(resolvedOutput);
+  } else {
+    outputDir = safeMkdtempSync(join(getTmpDir(), "scalasem-"));
+  }
+  return join(outputDir, basename(requested));
+}
+
+function readScalasemReport(reportFile) {
+  if (!reportFile || !safeExistsSync(reportFile)) {
+    return undefined;
+  }
+  try {
+    const parsed = JSON.parse(readFileSync(reportFile, "utf-8"));
+    return parsed && typeof parsed === "object" ? parsed : undefined;
+  } catch (_err) {
+    return undefined;
+  }
+}
+
+function isNewerThan(file, otherFile) {
+  try {
+    return statSync(file).mtimeMs >= statSync(otherFile).mtimeMs;
+  } catch (_err) {
+    // Without both timestamps the project checks alone decide.
+    return true;
+  }
+}
+
+/**
+ * A version 1 semantics slice (the shape atom's scalasem wrote before schema
+ * version 2): keyed by .scala files with `usedTypes`, no `_meta.schemaVersion`.
+ *
+ * @param {Object} parsed Candidate slice.
+ * @returns {boolean}
+ */
+export function isV1SemanticsSlice(parsed) {
+  if (!parsed || parsed._meta?.schemaVersion) {
+    return false;
+  }
+  return Object.keys(parsed).some((key) => key.endsWith(".scala"));
+}
+
+/**
+ * Whether a version 1 slice describes this project: it names no project, so
+ * every source file it lists must exist under the project directory.
+ */
+function isV1SliceOfProject(slice, projectDir) {
+  const files = Object.keys(slice).filter((key) => key.endsWith(".scala"));
+  return (
+    files.length > 0 &&
+    files.every((file) => {
+      const path = resolve(projectDir, file);
+      return path.startsWith(`${projectDir}${sep}`) && safeExistsSync(path);
+    })
+  );
+}
+
+/**
+ * Run scalasem over a Scala project and return its report, reusing a
+ * persisted version 2 report that belongs to this project and is newer than
+ * the input BOM.
+ *
+ * Failures are visible rather than silent: the reason is printed once, the
+ * run's diagnostics reach the caller as `cdx:scalasem:diagnostic` properties,
+ * and --fail-on-error claims the exit status. A version 1 slice of this
+ * project, at the path the user named or at the report path, is returned
+ * separately and never overwritten; the fresh report then goes to a
+ * temporary file.
+ *
+ * @param {string} src Project directory.
+ * @param {Object} options CLI options.
+ * @returns {Object|undefined} `{ report, reportFile, v1Slice, metadataProperties }`,
+ *   or undefined when the analyzer is disabled.
+ */
+export function analyzeScalaProject(src, options = {}) {
+  if (scalasemDisabled(options)) {
+    console.log(
+      "scalasem: disabled; skipping the Scala semantic evidence. Unset CDXGEN_SCALASEM_DISABLE or drop --no-scalasem to re-enable it.",
+    );
+    return undefined;
+  }
+  const analysisDir = resolvePluginSourceDir(src, SCALA_BUILD_MARKERS);
+  const reportFile = scalasemOutputFile(options);
+  const candidates = [
+    ...new Set(
+      [
+        options.semanticsSlicesFile
+          ? resolve(options.semanticsSlicesFile)
+          : undefined,
+        reportFile,
+      ].filter(Boolean),
+    ),
+  ];
+  let v1Slice;
+  let v1File;
+  for (const candidate of candidates) {
+    const existing = readScalasemReport(candidate);
+    if (existing?._meta?.schemaVersion === "scalasem/2") {
+      if (
+        resolve(existing._meta.projectPath || "") === analysisDir &&
+        isNewerThan(candidate, options.input)
+      ) {
+        if (DEBUG_MODE) {
+          console.log(`Reusing the scalasem report "${candidate}".`);
+        }
+        return {
+          report: existing,
+          reportFile: candidate,
+          v1Slice: undefined,
+          metadataProperties: scalasemMetadataProperties(existing),
+        };
+      }
+    } else if (
+      !v1Slice &&
+      isV1SemanticsSlice(existing) &&
+      isV1SliceOfProject(existing, analysisDir)
+    ) {
+      v1Slice = existing;
+      v1File = candidate;
+    }
+  }
+  return runScalasem(analysisDir, options, reportFile, v1Slice, v1File);
+}
+
+function scalasemRunFailure(code, message, options, args, status) {
+  console.warn(`scalasem: ${message}`);
+  deferFailOnError(options, {
+    ecosystem: "scala",
+    tool: "scalasem",
+    detail: message,
+    exitCode: typeof status === "number" ? status : undefined,
+    command: `${process.execPath} ${args.join(" ")}`,
+  });
+  return [{ code }];
+}
+
+function runScalasem(analysisDir, options, requestedFile, v1Slice, v1File) {
+  const scalasem = options.scalasemCommand || resolveScalasemCommand();
+  if (!scalasem) {
+    const diagnostics = scalasemRunFailure(
+      "scalasem-missing",
+      "analyzer not found, so the BOM carries no Scala semantic evidence. Install @appthreat/atom-parsetools or set SCALASEM_CMD.",
+      options,
+      [],
+    );
+    return {
+      report: undefined,
+      reportFile: undefined,
+      v1Slice,
+      metadataProperties: scalasemMetadataProperties(undefined, diagnostics),
+    };
+  }
+  // scalasem writes into a directory of this run's own, so a report found
+  // there afterwards is this run's. Writing to the requested path directly
+  // left only the file's modification time to tell a fresh report from one an
+  // earlier run (or another project's run) had left there, and a report
+  // written within the millisecond the run started passed for fresh.
+  const runDir = safeMkdtempSync(join(getTmpDir(), "scalasem-"));
+  const runReportFile = join(runDir, "scalasem.slices.json");
+  const args = [scalasem, analysisDir, runReportFile];
+  if (options.installDeps === false) {
+    args.push("--no-build");
+  }
+  if (options.scalasemIncludeTests && !options.requiredOnly) {
+    args.push("--include-tests");
+  }
+  if (DEBUG_MODE) {
+    console.log("Executing", process.execPath, args.join(" "));
+  }
+  const timeoutMs = scalasemTimeoutMs(options);
+  const startedAt = Date.now();
+  // scalasem enforces the limit itself, so the builds it started stop with
+  // it; cdxgen's own timeout is the backstop. Naming cdxgen as the supervisor
+  // stops scalasem when cdxgen is killed.
+  const result = safeSpawnSync(process.execPath, args, {
+    cwd: analysisDir,
+    shell: false,
+    killSignal: "SIGKILL",
+    timeout: timeoutMs + Math.min(TIMEOUT_GRACE_MS, timeoutMs),
+    env: {
+      ...process.env,
+      ATOM_PARENT_PID: String(process.pid),
+      SCALASEM_TIMEOUT: String(timeoutMs),
+    },
+  });
+  const elapsedMs = Date.now() - startedAt;
+  const wroteReport = safeExistsSync(runReportFile);
+  const written = wroteReport ? readScalasemReport(runReportFile) : undefined;
+  const usable = written?._meta?.schemaVersion === "scalasem/2";
+  const report = usable ? written : undefined;
+  // scalasem stops itself at the limit and exits with an error and no
+  // report; cdxgen's own timeout is only the backstop. A run that ends without
+  // a usable report once the limit has passed overran it, however it exited.
+  const timedOut =
+    result?.error?.code === "ETIMEDOUT" || (!usable && elapsedMs >= timeoutMs);
+  let diagnostics;
+  if (timedOut) {
+    diagnostics = scalasemRunFailure(
+      "scalasem-timeout",
+      `did not finish within ${formatDuration(timeoutMs)} and was stopped, so the BOM carries no Scala semantic evidence. Allow more time with CDXGEN_SCALASEM_TIMEOUT (milliseconds).`,
+      options,
+      args,
+    );
+  } else if (result?.error) {
+    diagnostics = scalasemRunFailure(
+      "scalasem-not-runnable",
+      `could not be started (${result.error.message}). Check SCALASEM_CMD or the atom-parsetools install.`,
+      options,
+      args,
+    );
+  } else if (!wroteReport) {
+    diagnostics = scalasemRunFailure(
+      "scalasem-no-report",
+      `wrote no report for ${analysisDir} (exit status ${result?.status ?? "unknown"}); run with CDXGEN_DEBUG_MODE=debug to see its output.`,
+      options,
+      args,
+      result?.status,
+    );
+  } else if (!written) {
+    diagnostics = scalasemRunFailure(
+      "scalasem-invalid-report",
+      `wrote a report for ${analysisDir} that is not valid JSON; run with CDXGEN_DEBUG_MODE=debug to see its output.`,
+      options,
+      args,
+      result?.status,
+    );
+  } else if (!usable) {
+    diagnostics = scalasemRunFailure(
+      "scalasem-old-report",
+      "wrote a version 1 report. Scala evidence needs atom-parsetools 1.10.0 or later.",
+      options,
+      args,
+      result?.status,
+    );
+  } else {
+    diagnostics = report._meta.diagnostics || [];
+    if (result?.status !== 0) {
+      diagnostics = [
+        ...diagnostics,
+        ...scalasemRunFailure(
+          "scalasem-exit-status",
+          `exited with status ${result?.status}; its report is used, and may be incomplete.`,
+          options,
+          args,
+          result?.status,
+        ),
+      ];
+    }
+    const summary = summarizeDiagnostics(report._meta.diagnostics);
+    if (summary) {
+      console.warn(
+        `scalasem: the report is degraded: ${summary}. The cdx:scalasem:diagnostic:* metadata properties of the BOM say where.`,
+      );
+    }
+  }
+  if (DEBUG_MODE && !usable && (result?.stdout || result?.stderr)) {
+    console.log(result.stdout, result.stderr);
+  }
+  // Only a usable report replaces the one at the requested path, and a
+  // version 1 slice there stays as the user left it.
+  let reportFile;
+  if (usable && v1File !== requestedFile) {
+    try {
+      safeMkdirSync(dirname(requestedFile), { recursive: true });
+      safeCopyFileSync(runReportFile, requestedFile);
+      reportFile = requestedFile;
+    } catch (err) {
+      console.warn(
+        `scalasem: the report could not be kept at ${requestedFile} (${err.message}); this run still uses it.`,
+      );
+    }
+  }
+  safeRmSync(runDir, { recursive: true, force: true });
+  return {
+    report,
+    reportFile,
+    v1Slice,
+    metadataProperties: scalasemMetadataProperties(report, diagnostics),
+  };
+}
+
+function summarizeDiagnostics(diagnostics = []) {
+  const byCode = new Map();
+  for (const diagnostic of diagnostics || []) {
+    if (diagnostic?.code) {
+      byCode.set(
+        diagnostic.code,
+        (byCode.get(diagnostic.code) || 0) + (diagnostic.count ?? 1),
+      );
+    }
+  }
+  return [...byCode.entries()]
+    .sort(([left], [right]) => left.localeCompare(right))
+    .map(([code, count]) => `${code} (${count})`)
+    .join(", ");
+}
+
+/**
+ * The `cdx:scalasem:*` metadata properties of one run: what produced the
+ * report, and what limited it.
+ *
+ * @param {Object|undefined} report Parsed scalasem report, when one exists.
+ * @param {Object[]} diagnostics Run diagnostics, from the report or the spawn.
+ * @returns {Object[]} Metadata properties.
+ */
+export function scalasemMetadataProperties(
+  report,
+  diagnostics = report?._meta?.diagnostics || [],
+) {
+  const properties = [];
+  const meta = report?._meta || {};
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:schemaVersion",
+    meta.schemaVersion,
+  );
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:factsSource",
+    sortedCsv(meta.generatedFrom),
+  );
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:compilerSource",
+    sortedCsv((meta.compilers || []).map((compiler) => compiler.source)),
+  );
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:scalaVersions",
+    sortedCsv((meta.compilers || []).map((compiler) => compiler.version)),
+  );
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:platforms",
+    sortedCsv(meta.platforms),
+  );
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:filesAnalyzed",
+    meta.counts?.files,
+  );
+  if (meta.truncated || diagnostics.length) {
+    appendUniqueProperty(properties, "cdx:scalasem:degraded", "true");
+  }
+  const byCode = {};
+  for (const diagnostic of diagnostics) {
+    if (!diagnostic?.code) {
+      continue;
+    }
+    byCode[diagnostic.code] =
+      (byCode[diagnostic.code] || 0) + (diagnostic.count ?? 1);
+  }
+  for (const code of Object.keys(byCode).sort()) {
+    appendUniqueProperty(
+      properties,
+      `cdx:scalasem:diagnostic:${code}`,
+      byCode[code],
+    );
+  }
+  return properties;
+}
+
+export function sortedCsv(values) {
+  const items = values instanceof Set ? [...values] : values || [];
+  const filtered = [...new Set(items.filter(Boolean))].sort();
+  return filtered.length ? filtered.join(",") : undefined;
+}
+
+/**
+ * The platform a purl's artifact name targets: `_sjs1` is Scala.js,
+ * `_native0.x` is Scala Native, and neither is the JVM.
+ *
+ * @param {string} purl Package URL.
+ * @returns {"jvm"|"js"|"native"} Platform key.
+ */
+export function scalaPurlPlatform(purl) {
+  const name = purlPurlName(purl);
+  if (!name) {
+    return "jvm";
+  }
+  if (/_(sjs\d+(?:\.\d+)?)$/.test(name)) {
+    return "js";
+  }
+  if (/_(native\d+(?:\.\d+)*)$/.test(name)) {
+    return "native";
+  }
+  return "jvm";
+}
+
+function purlPurlName(purl) {
+  const match = /pkg:[^/]+\/(?:[^/@?]+\/)?([^/@?]+)/.exec(purl || "");
+  if (!match) {
+    return undefined;
+  }
+  try {
+    return decodeURIComponent(match[1]);
+  } catch (_err) {
+    return match[1];
+  }
+}
+
+/**
+ * The class and package index that joins scalasem symbols to component purls.
+ *
+ * Symbols are indexed from three sources in priority order: the classpath the
+ * report recorded (1), the `internal:Namespaces` properties of the components
+ * (2), and the jar namespace map cdxgen writes beside the BOM (3). A symbol an
+ * earlier source already attributed is never re-attributed by a later one.
+ *
+ * A jar answers for the classes it ships and for the packages it ships classes
+ * in directly, never for the parents of those packages: a library with one
+ * class under `scala.collection.compat` owns neither `scala.collection` nor
+ * `scala`. The JDK has no component, and neither do the project's own classes.
+ */
+export class ScalaJoinIndex {
+  constructor() {
+    this.classPurls = new Map();
+    this.packagePurls = new Map();
+    this.sourceBySymbol = new Map();
+    this.purlsBySource = new Map();
+    this.namespacesByPurl = new Map();
+    this.projectClasses = new Set();
+    this.projectPackages = new Set();
+  }
+
+  hasNamespaces(purl) {
+    return (this.purlsBySource.get(purl) || 0) > 0;
+  }
+
+  /**
+   * Index the classes of one library: each class under its dotted name (a
+   * nested `Outer$Inner` also as `Outer.Inner`, an object `Foo$` as `Foo`),
+   * and the package each one sits in.
+   *
+   * @param {string} purl The owning component purl.
+   * @param {string[]} names Class names from the jar or the namespace property.
+   * @param {number} source The source priority, 1 to 3.
+   */
+  addNamespaces(purl, names, source) {
+    if (!purl || this.hasNamespaces(purl)) {
+      return;
+    }
+    this.purlsBySource.set(purl, source);
+    const indexed = [];
+    this.namespacesByPurl.set(purl, indexed);
+    for (const name of names || []) {
+      const symbol = String(name || "").trim();
+      if (!symbol || symbol.includes(" ") || symbol.includes("/")) {
+        continue;
+      }
+      indexed.push(symbol);
+      const dollar = symbol.indexOf("$");
+      const topLevel = dollar > 0 ? symbol.slice(0, dollar) : symbol;
+      this.addSymbol(this.classPurls, topLevel, purl, source);
+      if (dollar > 0) {
+        const nested = symbol.replaceAll("$", ".").replace(/\.+$/, "");
+        if (nested !== topLevel) {
+          this.addSymbol(this.classPurls, nested, purl, source);
+        }
+      }
+      const lastDot = topLevel.lastIndexOf(".");
+      if (lastDot > 0) {
+        this.addSymbol(
+          this.packagePurls,
+          topLevel.slice(0, lastDot),
+          purl,
+          source,
+        );
+      }
+    }
+  }
+
+  addSymbol(map, symbol, purl, source) {
+    const key = `${map === this.classPurls ? "c" : "p"}:${symbol}`;
+    const existing = this.sourceBySymbol.get(key);
+    if (existing !== undefined && existing < source) {
+      // A higher-priority source (lower number) already attributed it.
+      return;
+    }
+    if (existing !== undefined && existing > source) {
+      map.set(symbol, new Set());
+    }
+    this.sourceBySymbol.set(key, source);
+    if (!map.has(symbol)) {
+      map.set(symbol, new Set());
+    }
+    map.get(symbol).add(purl);
+  }
+
+  /**
+   * Record the classes the project defines, so that references to them, and
+   * to packages only the project fills, never join a library.
+   *
+   * @param {Object} report Parsed scalasem report.
+   */
+  addProjectDefinitions(report) {
+    for (const [fileKey, fileEntry] of Object.entries(report || {})) {
+      if (!isReportFileKey(fileKey) || !Array.isArray(fileEntry?.definitions)) {
+        continue;
+      }
+      for (const definition of fileEntry.definitions) {
+        if (
+          !["class", "object", "trait", "enum"].includes(definition?.kind) ||
+          !definition.owner ||
+          !definition.name
+        ) {
+          continue;
+        }
+        this.projectClasses.add(`${definition.owner}.${definition.name}`);
+        this.projectPackages.add(definition.owner);
+      }
+    }
+  }
+
+  /**
+   * Whether a symbol belongs to code the BOM has no component for: the JDK,
+   * or the project itself.
+   */
+  isUnowned(symbol) {
+    if (JDK_OWNER_PREFIXES.some((prefix) => symbol.startsWith(prefix))) {
+      return true;
+    }
+    const parts = symbol.replace(/\$+$/, "").split(".");
+    for (let i = parts.length; i > 1; i--) {
+      if (this.projectClasses.has(parts.slice(0, i).join("."))) {
+        return true;
+      }
+    }
+    return false;
+  }
+
+  /**
+   * The libraries that ship the class a symbol names: the longest dotted
+   * prefix of the symbol that is a class name, for a member or a nested type.
+   */
+  classOwners(symbol) {
+    const parts = symbol.replace(/\$+$/, "").split(".");
+    for (let i = parts.length; i > 0; i--) {
+      const owners = this.classPurls.get(parts.slice(0, i).join("."));
+      if (owners?.size) {
+        return owners;
+      }
+    }
+    return undefined;
+  }
+
+  /**
+   * The libraries that ship classes directly in the package a symbol names or
+   * sits in. A package the project also fills is ambiguous and joins nothing,
+   * and `javax.` names join only through an exact class, since the JDK ships
+   * most of them.
+   */
+  packageOwners(symbol) {
+    if (symbol.startsWith(JDK_EXTENSION_PREFIX)) {
+      return undefined;
+    }
+    const parts = symbol.replace(/\$+$/, "").split(".");
+    for (let i = parts.length; i > 0; i--) {
+      const name = parts.slice(0, i).join(".");
+      const owners = this.packagePurls.get(name);
+      if (owners?.size) {
+        return this.projectPackages.has(name) ? undefined : owners;
+      }
+    }
+    return undefined;
+  }
+
+  /**
+   * The purl owning a symbol: the class first, then the package it sits in,
+   * each narrowed to the module's own classpath and the file's platform, and
+   * kept only when a single owner remains.
+   *
+   * @param {string} symbol Fully qualified symbol from the report.
+   * @param {string|undefined} platform Platform of the referencing file.
+   * @param {Set<string>|undefined} allowedPurls Purls on the module's classpath.
+   * @returns {string|undefined} Component purl.
+   */
+  lookup(symbol, platform, allowedPurls) {
+    if (!symbol || this.isUnowned(symbol)) {
+      return undefined;
+    }
+    const byClass = this.narrow(
+      this.classOwners(symbol),
+      platform,
+      allowedPurls,
+    );
+    if (byClass.length) {
+      return byClass.length === 1 ? byClass[0] : undefined;
+    }
+    const byPackage = this.narrow(
+      this.packageOwners(symbol),
+      platform,
+      allowedPurls,
+    );
+    return byPackage.length === 1 ? byPackage[0] : undefined;
+  }
+
+  /**
+   * The owners a stack is evidence for: every jar that ships the class, since
+   * which one served the call cannot be told from the source, or the single
+   * package owner when no jar lists the class.
+   */
+  lookupAll(symbol, platform, allowedPurls) {
+    if (!symbol || this.isUnowned(symbol)) {
+      return [];
+    }
+    const byClass = this.narrow(
+      this.classOwners(symbol),
+      platform,
+      allowedPurls,
+    );
+    if (byClass.length) {
+      return byClass;
+    }
+    const single = this.lookup(symbol, platform, allowedPurls);
+    return single ? [single] : [];
+  }
+
+  /**
+   * Narrow candidate owners to what the referencing file can use. A module
+   * compiles against its own classpath: a library on it wins, and a library
+   * only other modules compile against does not qualify. A library on no
+   * module's classpath stays a candidate, since the report cannot list every
+   * jar (a locally published snapshot, say). The file's platform then picks
+   * among the JVM, Scala.js and Scala Native variants of one library.
+   */
+  narrow(candidates, platform, allowedPurls) {
+    let pool = [...(candidates || [])];
+    if (!pool.length) {
+      return pool;
+    }
+    let onClasspath = false;
+    if (allowedPurls?.size) {
+      const inModule = pool.filter((purl) => allowedPurls.has(purl));
+      onClasspath = inModule.length > 0;
+      pool = onClasspath
+        ? inModule
+        : pool.filter((purl) => !this.classpathPurls?.has(purl));
+    }
+    if (platform && pool.length) {
+      const samePlatform = pool.filter(
+        (purl) => scalaPurlPlatform(purl) === platform,
+      );
+      if (samePlatform.length || !onClasspath) {
+        pool = samePlatform;
+      }
+    } else if (pool.length > 1) {
+      // Without a platform the JVM speaks for a library built for several.
+      const jvm = pool.filter((purl) => scalaPurlPlatform(purl) === "jvm");
+      if (jvm.length) {
+        pool = jvm;
+      }
+    }
+    return pool;
+  }
+}
+
+/**
+ * Whether a report key names a source file entry. The schema keys file
+ * entries by their relative path, for Scala sources, scripts and the Java
+ * sources of a mixed module alike.
+ */
+export function isReportFileKey(key) {
+  return /\.(scala|sc|java)$/.test(key || "");
+}
+
+function decodeSafe(value) {
+  if (!value) {
+    return "";
+  }
+  try {
+    return decodeURIComponent(value);
+  } catch (_err) {
+    return value;
+  }
+}
+
+// The Scala binary versions a component was built for: the suffix its purl
+// keeps, or every cdx:scala:compilerVersion it carries, since one component can
+// stand for the same library in builds of several Scala versions.
+function componentBinaryVersions(component, coordinate) {
+  if (coordinate.binaryVersion) {
+    return new Set([coordinate.binaryVersion]);
+  }
+  return new Set(
+    (component.properties || [])
+      .filter((property) => property.name === "cdx:scala:compilerVersion")
+      .map((property) => property.value)
+      .filter(Boolean),
+  );
+}
+
+/**
+ * Index the BOM components by their normalized Scala coordinates, both with
+ * and without the version, so a classpath jar, a namespace property and a map
+ * entry of the same library meet on one component. Each entry keeps the Scala
+ * binary version the component was built for, so the `_2.13` and `_3`
+ * artifacts of one library in a Maven build stay apart.
+ */
+export function componentCoordinateIndex(components = []) {
+  const byVersion = new Map();
+  const withoutVersion = new Map();
+  for (const component of components) {
+    const coordinate = scalaCoordinateOfPurl(component?.purl);
+    if (!coordinate) {
+      continue;
+    }
+    const [type, group, artifactBase, platformSuffix] =
+      coordinate.key.split(":");
+    const entry = {
+      component,
+      binaryVersions: componentBinaryVersions(component, coordinate),
+    };
+    if (!byVersion.has(coordinate.key)) {
+      byVersion.set(coordinate.key, []);
+    }
+    byVersion.get(coordinate.key).push(entry);
+    const key = scalaCoordinateKey({
+      type,
+      group,
+      artifactBase,
+      platformSuffix,
+    });
+    if (!withoutVersion.has(key)) {
+      withoutVersion.set(key, []);
+    }
+    withoutVersion.get(key).push(entry);
+  }
+  return { byVersion, withoutVersion };
+}
+
+function pickByBinaryVersion(entries, binaryVersion) {
+  if (!entries?.length) {
+    return [];
+  }
+  if (binaryVersion) {
+    const same = entries.filter((entry) =>
+      entry.binaryVersions.has(binaryVersion),
+    );
+    if (same.length) {
+      return same;
+    }
+    const unknown = entries.filter((entry) => !entry.binaryVersions.size);
+    if (unknown.length) {
+      return unknown;
+    }
+    return [];
+  }
+  return entries;
+}
+
+/**
+ * The component a report classpath entry belongs to. Coordinates are
+ * normalized (`upickle_sjs1_3` and the purl name `upickle_sjs1` share a key),
+ * the entry's version and Scala binary version win, and a versionless match
+ * must have a single owner. A jar the BOM does not hold matches nothing: no
+ * component is invented.
+ *
+ * @returns {{component: Object, exact: boolean}|undefined} The component, and
+ *   whether its version is the one on the classpath.
+ */
+export function matchClasspathEntry(
+  entry,
+  { byVersion, withoutVersion },
+  platform,
+) {
+  const parsed = parseScalaArtifact(entry?.artifact);
+  const parts = {
+    group: entry?.group,
+    artifactBase: parsed.artifactBase,
+    platformSuffix: parsed.platformSuffix,
+  };
+  const exact = pickByBinaryVersion(
+    byVersion.get(scalaCoordinateKey({ ...parts, version: entry?.version })),
+    parsed.binaryVersion,
+  );
+  if (exact.length === 1) {
+    return { component: exact[0].component, exact: true };
+  }
+  if (exact.length > 1) {
+    return undefined;
+  }
+  const candidates = pickByBinaryVersion(
+    withoutVersion.get(scalaCoordinateKey(parts)),
+    parsed.binaryVersion,
+  );
+  const matching = candidates.filter(
+    (candidate) => scalaPurlPlatform(candidate.component.purl) === platform,
+  );
+  const pool = matching.length ? matching : candidates;
+  return pool.length === 1
+    ? { component: pool[0].component, exact: false }
+    : undefined;
+}
+
+/**
+ * Build the symbol index from the three sources, reading each classpath jar
+ * once, and only for components that carry no namespaces yet.
+ *
+ * @param {Object} report Parsed scalasem report.
+ * @param {Object[]} components BOM components.
+ * @param {Object} jarNSMapping The jar namespace map cdxgen writes beside the BOM.
+ * @returns {Promise<ScalaJoinIndex>}
+ */
+export async function buildScalaJoinIndex(report, components, jarNSMapping) {
+  const index = new ScalaJoinIndex();
+  index.addProjectDefinitions(report);
+  index.exactPurls = new Set();
+  const coordinateIndex = componentCoordinateIndex(components);
+  const purlsWithNamespaces = new Set(
+    components
+      .filter((component) =>
+        (component.properties || []).some(
+          (property) => property.name === "internal:Namespaces",
+        ),
+      )
+      .map((component) => component.purl),
+  );
+  const readJars = new Set();
+  for (const module of report.modules || []) {
+    for (const entry of module.classpath || []) {
+      const matched = matchClasspathEntry(
+        entry,
+        coordinateIndex,
+        module.platform,
+      );
+      const purl = matched?.component?.purl;
+      if (!purl || purlsWithNamespaces.has(purl) || index.hasNamespaces(purl)) {
+        continue;
+      }
+      if (!entry.path || readJars.has(entry.path)) {
+        continue;
+      }
+      readJars.add(entry.path);
+      if (!safeExistsSync(entry.path)) {
+        continue;
+      }
+      let classes = [];
+      try {
+        classes = await getJarClasses(entry.path);
+      } catch (_err) {
+        classes = [];
+      }
+      if (classes.length) {
+        index.addNamespaces(purl, classes, 1);
+        if (matched.exact) {
+          index.exactPurls.add(purl);
+        }
+      }
+    }
+  }
+  for (const component of components) {
+    if (purlsWithNamespaces.has(component.purl)) {
+      const namespaces = (component.properties || [])
+        .filter((property) => property.name === "internal:Namespaces")
+        .flatMap((property) => (property.value || "").split("\n"));
+      index.addNamespaces(component.purl, namespaces, 2);
+    }
+  }
+  for (const mapPurl of Object.keys(jarNSMapping || {})) {
+    const entry = jarNSMapping[mapPurl] || {};
+    if (!Array.isArray(entry.namespaces) || !entry.namespaces.length) {
+      continue;
+    }
+    const segments = /^pkg:[^/]+\/(?:([^/@?]+)\/)?([^@?]+)@([^?#]*)/.exec(
+      mapPurl,
+    );
+    if (!segments) {
+      continue;
+    }
+    const matched = matchClasspathEntry(
+      {
+        group: decodeSafe(segments[1]),
+        artifact: decodeSafe(segments[2]),
+        version: decodeSafe(segments[3]),
+      },
+      coordinateIndex,
+      scalaPurlPlatform(mapPurl),
+    );
+    if (matched) {
+      index.addNamespaces(matched.component.purl, entry.namespaces, 3);
+    }
+  }
+  return index;
+}
+
+// The registry of known algorithm OIDs, shared with the golem converter.
+const SCALASEM_CRYPTO_OIDS = JSON.parse(
+  readFileSync(join(dirNameStr, "data", "crypto-oid.json"), "utf-8"),
+);
+const SCALASEM_OID_KEYS = new Map(
+  Object.keys(SCALASEM_CRYPTO_OIDS).map((key) => [key.toLowerCase(), key]),
+);
+
+// Canonical scalasem algorithm spellings that have no key of their own in the
+// OID registry, mapped to the key that names the same algorithm. Everything
+// the rules can emit is covered by this table, the key-size dependent AES
+// family below, or one of the explicit no-OID sets.
+const SCALASEM_OID_ALIASES = {
+  HS256: "hmacWithSHA256",
+  HS384: "hmacWithSHA384",
+  HS512: "hmacWithSHA512",
+  HmacSHA1: "hmacWithSHA1",
+  HmacSHA224: "hmacWithSHA224",
+  HmacSHA256: "hmacWithSHA256",
+  HmacSHA384: "hmacWithSHA384",
+  HmacSHA512: "hmacWithSHA512",
+  ES256: "ecdsaWithSHA256",
+  ES384: "ecdsaWithSHA384",
+  ES512: "ecdsaWithSHA512",
+  RS256: "sha256WithRSAEncryption",
+  RS384: "sha384WithRSAEncryption",
+  RS512: "sha512WithRSAEncryption",
+  PS256: "rsassa-pss",
+  PS384: "rsassa-pss",
+  PS512: "rsassa-pss",
+  SHA1withDSA: "dsaWithSha1",
+  SHA1withRSA: "sha1-with-rsa-signature",
+  SHA224withDSA: "dsaWithSha224",
+  SHA256withDSA: "dsaWithSha256",
+  SHA1withECDSA: "ecdsaWithSHA1",
+  SHA224withECDSA: "ecdsaWithSHA224",
+  SHA256withECDSA: "ecdsaWithSHA256",
+  SHA384withECDSA: "ecdsaWithSHA384",
+  SHA512withECDSA: "ecdsaWithSHA512",
+  SHA224withRSA: "sha224WithRSAEncryption",
+  SHA256withRSA: "sha256WithRSAEncryption",
+  SHA384withRSA: "sha384WithRSAEncryption",
+  SHA512withRSA: "sha512WithRSAEncryption",
+  MD2withRSA: "md2WithRSAEncryption",
+  MD5withRSA: "md5WithRSAEncryption",
+  "RSA-OAEP": "id-RSAES-OAEP",
+  "RSASSA-PKCS1-v1_5": "rsaSignature",
+  "SHA256withRSA/PSS": "rsassa-pss",
+  "ChaCha20-Poly1305": "chaCha20Poly1305",
+  X448: "curveX448",
+  PBKDF2WithHmacSHA1: "PBKDF2",
+  PBKDF2WithHmacSHA256: "PBKDF2",
+  PBKDF2WithHmacSHA384: "PBKDF2",
+  PBKDF2WithHmacSHA512: "PBKDF2",
+  PBEWithMD5AndDES: "pbeWithMD5AndDES-CBC",
+};
+
+// AES spellings resolve through the key size and mode when both are known,
+// and to the generic `aes` key otherwise.
+const SCALASEM_AES_MODE_KEYS = {
+  ECB: "ECB",
+  CBC: "CBC",
+  CFB: "CFB",
+  OFB: "OFB",
+  CTR: "CTR",
+  GCM: "GCM",
+  CCM: "CCM",
+  GMAC: "GMAC",
+};
+const SCALASEM_AES_NAMES = new Set([
+  "AES",
+  "AES-CBC",
+  "AES-CTR",
+  "AES-GCM",
+  "AES-KW",
+]);
+
+// Names that never resolve to an OID. Protocols, keystores and random
+// generators are emitted as their own asset kinds without an OID; the
+// algorithms listed here have no registry key and become
+// `cdx:scalasem:crypto:*` properties instead of assets. Ciphers whose key
+// depends on the mode (Blowfish, DESede, RC2) and the curve algorithms
+// resolve only when the finding names the mode or the curve, and are
+// properties otherwise.
+export const SCALASEM_PROTOCOL_NAMES = new Set([
+  "SSL",
+  "SSLv3",
+  "TLS",
+  "TLSv1",
+  "TLSv1.1",
+  "TLSv1.2",
+  "TLSv1.3",
+  "DTLSv1.2",
+]);
+export const SCALASEM_KEYSTORE_NAMES = new Set([
+  "JKS",
+  "JCEKS",
+  "PKCS12",
+  "BCFKS",
+]);
+// Algorithms that resolve to an OID only together with what the finding
+// says about them: the mode of a mode-keyed cipher, or the curve.
+export const SCALASEM_QUALIFIED_ALGORITHMS = new Set([
+  "Blowfish",
+  "DESede",
+  "EC",
+  "ECDH",
+  "ECDSA",
+  "ECIES",
+  "EdDSA",
+  "RC2",
+  "XDH",
+]);
+export const SCALASEM_NO_OID_ALGORITHMS = new Set([
+  "Argon2",
+  "Argon2d",
+  "Argon2i",
+  "Argon2id",
+  "BLAKE2b-512",
+  "BLAKE2s-256",
+  "CMAC",
+  "Camellia",
+  "ChaCha20",
+  "DH",
+  "DRBG",
+  "HKDF",
+  "HMAC",
+  "HmacSHA3-256",
+  "HmacSHA3-512",
+  "MD2withDSA",
+  "MD2withECDSA",
+  "MD5withDSA",
+  "MD5withECDSA",
+  "NativePRNG",
+  "NONEwithECDSA",
+  "NONEwithRSA",
+  "PBEWithHmacSHA256AndAES_128",
+  "PBEWithHmacSHA256AndAES_256",
+  "PBEWithMD5AndTripleDES",
+  "PBEWithSHA1AndDESede",
+  "PBEWithSHA1AndRC4_128",
+  "Poly1305",
+  "SHA-512/224",
+  "SHA-512/256",
+  "SHA1PRNG",
+  "SHA3-256withECDSA",
+  "SHA384withDSA",
+  "SHA512withDSA",
+  "SM3",
+  "SM4",
+  "Whirlpool",
+  "Windows-PRNG",
+  "bcrypt",
+  "none",
+  "scrypt",
+]);
+
+// Ciphers whose registry key names the mode: `blowfishCBC`, `rc2ECB`.
+const SCALASEM_MODE_KEYED_CIPHERS = {
+  Blowfish: (mode) => `blowfish${mode}`,
+  DESede: (mode) => (mode === "CBC" ? "des-EDE3-CBC" : undefined),
+  RC2: (mode) => `rc2${mode}`,
+};
+
+// NIST names of the prime curves, as the registry spells them.
+const SCALASEM_CURVE_ALIASES = {
+  "p-256": "secp256r1",
+  "p-384": "secp384r1",
+  "p-521": "secp521r1",
+};
+const SCALASEM_WEIERSTRASS_ALGORITHMS = new Set([
+  "EC",
+  "ECDH",
+  "ECDSA",
+  "ECIES",
+]);
+
+/**
+ * The registry key of a curve algorithm on a named curve. A key pair or an
+ * ECDSA or ECDH use on a prime curve is identified by that curve. The
+ * Montgomery curves serve key agreement only and the Edwards curves
+ * signatures only, so a curve the algorithm cannot use resolves to nothing.
+ */
+function curveOidKey(name, curve) {
+  const normalized = String(curve || "").toLowerCase();
+  if (!normalized) {
+    return undefined;
+  }
+  const montgomery = /^(x|curve)(25519|448)$/.exec(normalized);
+  const edwards = /^ed(25519|448)$/.exec(normalized);
+  if (montgomery) {
+    return ["XDH", "ECDH"].includes(name)
+      ? SCALASEM_OID_KEYS.get(
+          montgomery[2] === "25519" ? "x25519" : "curvex448",
+        )
+      : undefined;
+  }
+  if (edwards) {
+    return name === "EdDSA"
+      ? SCALASEM_OID_KEYS.get(`ed${edwards[1]}`)
+      : undefined;
+  }
+  if (!SCALASEM_WEIERSTRASS_ALGORITHMS.has(name)) {
+    return undefined;
+  }
+  return SCALASEM_OID_KEYS.get(
+    SCALASEM_CURVE_ALIASES[normalized] || normalized,
+  );
+}
+
+/**
+ * The CycloneDX padding enum value of a JCA padding name.
+ */
+function paddingOf(padding) {
+  const normalized = String(padding || "").toLowerCase();
+  if (!normalized) {
+    return undefined;
+  }
+  if (normalized.startsWith("oaep")) {
+    return "oaep";
+  }
+  return SCALASEM_PADDING[normalized] || "other";
+}
+
+/**
+ * The OID registry key for a canonical scalasem algorithm name, with the key
+ * size, mode, curve or padding when the name needs one. Names that have no
+ * key return undefined.
+ *
+ * @param {string} name Canonical algorithm name.
+ * @param {number} [keySize] Key size in bits, when the report carries one.
+ * @param {string} [mode] Block cipher mode, such as `GCM`.
+ * @param {string} [curve] Elliptic curve name.
+ * @param {string} [padding] JCA padding name.
+ * @returns {string|undefined} Registry key such as `aes256-GCM`.
+ */
+export function scalasemCryptoOidKey(name, keySize, mode, curve, padding) {
+  if (!name) {
+    return undefined;
+  }
+  if (SCALASEM_AES_NAMES.has(name)) {
+    const modeKey =
+      SCALASEM_AES_MODE_KEYS[mode] ||
+      SCALASEM_AES_MODE_KEYS[name.split("-")[1]];
+    const wrap = name === "AES-KW" || mode === "KW";
+    if (keySize && (modeKey !== undefined || wrap)) {
+      const sizedKey = wrap ? `aes${keySize}-wrap` : `aes${keySize}-${modeKey}`;
+      const hit = SCALASEM_OID_KEYS.get(sizedKey.toLowerCase());
+      if (hit) {
+        return hit;
+      }
+    }
+    return SCALASEM_OID_KEYS.get("aes");
+  }
+  if (SCALASEM_MODE_KEYED_CIPHERS[name]) {
+    const key = mode && SCALASEM_MODE_KEYED_CIPHERS[name](mode.toUpperCase());
+    return key ? SCALASEM_OID_KEYS.get(key.toLowerCase()) : undefined;
+  }
+  if (name === "RSA") {
+    const rsaPadding = paddingOf(padding);
+    if (rsaPadding === "oaep") {
+      return SCALASEM_OID_KEYS.get("id-rsaes-oaep");
+    }
+    if (rsaPadding === "pkcs1v15") {
+      return SCALASEM_OID_KEYS.get("rsaencryption");
+    }
+  }
+  const aliased = SCALASEM_OID_ALIASES[name];
+  if (aliased && SCALASEM_OID_KEYS.has(aliased.toLowerCase())) {
+    return SCALASEM_OID_KEYS.get(aliased.toLowerCase());
+  }
+  if (
+    SCALASEM_WEIERSTRASS_ALGORITHMS.has(name) ||
+    name === "XDH" ||
+    name === "EdDSA"
+  ) {
+    return curveOidKey(name, curve);
+  }
+  return SCALASEM_OID_KEYS.get(name.toLowerCase());
+}
+
+function maxOccurrences() {
+  const parsed = Number.parseInt(
+    String(readEnvironmentVariable("CDXGEN_SCALASEM_MAX_OCCURRENCES") ?? ""),
+    10,
+  );
+  return Number.isFinite(parsed) && parsed > 0
+    ? parsed
+    : DEFAULT_MAX_OCCURRENCES;
+}
+
+function isStdlibPurl(purl) {
+  return SCALA_STDLIB_ARTIFACTS.has(
+    parseScalaArtifact(purlPurlName(purl) || "").artifactBase,
+  );
+}
+
+/**
+ * The component purls each module's own classpath names, so a file compiled
+ * against one version of a library attributes its uses to that version.
+ */
+function classpathPurlsByModule(report, coordinateIndex) {
+  const byModule = new Map();
+  for (const module of report.modules || []) {
+    const purls = new Set();
+    for (const entry of module.classpath || []) {
+      const matched = matchClasspathEntry(
+        entry,
+        coordinateIndex,
+        module.platform,
+      );
+      if (matched) {
+        purls.add(matched.component.purl);
+      }
+    }
+    byModule.set(module.id, purls);
+  }
+  return byModule;
+}
+
+function allowedPurlsFor(fileEntry, allowedPurlsByModule) {
+  if (!fileEntry?.module) {
+    return undefined;
+  }
+  return allowedPurlsByModule.get(fileEntry.module);
+}
+
+/**
+ * Project a scalasem report onto CycloneDX evidence: occurrences for the
+ * joined components, call stacks for the library sinks, namespaces for the
+ * components the classpath join read, and the run's metadata properties.
+ *
+ * @param {Object} report Parsed scalasem report.
+ * @param {Object[]} components BOM components.
+ * @param {Object} options CLI options; `scalaNamespaceMap` carries the jar
+ *   namespace map cdxgen wrote beside the BOM.
+ * @returns {Promise<Object>} The evidence maps `createEvinseFile` consumes.
+ */
+export async function collectScalasemEvidence(
+  report = {},
+  components = [],
+  options = {},
+) {
+  const joinIndex = await buildScalaJoinIndex(
+    report,
+    components,
+    options.scalaNamespaceMap || {},
+  );
+  const purlLocationMap = {};
+  const componentPropertiesMap = {};
+  const dataFlowFrames = {};
+  const metadataProperties = scalasemMetadataProperties(report);
+  const cryptoComponentsByRef = new Map();
+  const cryptoGeneratePurls = {};
+  if (options.deep || options.scalasemNamespaces) {
+    attachJoinedNamespaces(joinIndex, components, componentPropertiesMap);
+  }
+  const allowedPurlsByModule = classpathPurlsByModule(
+    report,
+    componentCoordinateIndex(components),
+  );
+  joinIndex.classpathPurls = new Set(
+    [...allowedPurlsByModule.values()].flatMap((purls) => [...purls]),
+  );
+  addUsageEvidence(
+    report,
+    joinIndex,
+    purlLocationMap,
+    componentPropertiesMap,
+    options,
+    allowedPurlsByModule,
+  );
+  addCallStackEvidence(
+    report,
+    joinIndex,
+    dataFlowFrames,
+    allowedPurlsByModule,
+    options,
+  );
+  addCryptoEvidence(
+    report,
+    components,
+    cryptoComponentsByRef,
+    cryptoGeneratePurls,
+    componentPropertiesMap,
+    metadataProperties,
+    options,
+  );
+  addJsModuleEvidence(
+    report,
+    components,
+    purlLocationMap,
+    componentPropertiesMap,
+    metadataProperties,
+  );
+  return {
+    componentPropertiesMap,
+    cryptoComponents: Array.from(cryptoComponentsByRef.values()).sort(
+      (left, right) =>
+        `${left.name}:${left["bom-ref"]}`.localeCompare(
+          `${right.name}:${right["bom-ref"]}`,
+        ),
+    ),
+    cryptoGeneratePurls,
+    dataFlowFrames,
+    joinIndex,
+    metadataProperties,
+    purlLocationMap,
+  };
+}
+
+/**
+ * With --deep, publish the namespaces the classpath join read as
+ * `internal:Namespaces` properties, the property the deep Maven flow
+ * attaches, so components from builds that report no namespaces gain them.
+ * Only a jar of the component's own version speaks for its classes.
+ */
+function attachJoinedNamespaces(joinIndex, components, componentPropertiesMap) {
+  const known = new Set(
+    components
+      .filter((component) =>
+        (component.properties || []).some(
+          (property) => property.name === "internal:Namespaces",
+        ),
+      )
+      .map((component) => component.purl),
+  );
+  for (const [purl, source] of joinIndex.purlsBySource.entries()) {
+    if (source !== 1 || known.has(purl) || !joinIndex.exactPurls?.has(purl)) {
+      continue;
+    }
+    const namespaces = joinIndex.namespacesByPurl.get(purl) || [];
+    if (namespaces.length) {
+      componentPropertiesMap[purl] ??= [];
+      appendUniqueProperty(
+        componentPropertiesMap[purl],
+        "internal:Namespaces",
+        namespaces.join("\n"),
+      );
+    }
+  }
+}
+
+/**
+ * Occurrences from the report's references and calls: every symbol and call
+ * owner joined to a component records its source location. Test sources count
+ * only when they were asked for, the Scala runtime is capped at one
+ * occurrence per file, and each component keeps at most a bounded number of
+ * locations, every file it is used in first and further lines after. Call
+ * sites are counted in full, whatever the cap kept.
+ */
+function addUsageEvidence(
+  report,
+  joinIndex,
+  purlLocationMap,
+  componentPropertiesMap,
+  options,
+  allowedPurlsByModule,
+) {
+  const includeTests = options.scalasemIncludeTests === true;
+  const cap = maxOccurrences();
+  const scopesByPurl = {};
+  const platformsByPurl = {};
+  const callSitesByPurl = {};
+  const stdlibFilesByPurl = {};
+  const linesByPurl = new Map();
+  const addOccurrence = (purl, file, line, kind, scope, platform) => {
+    if (!purl || !line) {
+      return;
+    }
+    addSetValue(scopesByPurl, purl, scope);
+    addSetValue(platformsByPurl, purl, platform);
+    if (kind === "call") {
+      callSitesByPurl[purl] = (callSitesByPurl[purl] || 0) + 1;
+    }
+    if (isStdlibPurl(purl)) {
+      stdlibFilesByPurl[purl] ??= new Set();
+      if (stdlibFilesByPurl[purl].has(file)) {
+        return;
+      }
+      stdlibFilesByPurl[purl].add(file);
+    }
+    if (!linesByPurl.has(purl)) {
+      linesByPurl.set(purl, new Map());
+    }
+    const byFile = linesByPurl.get(purl);
+    if (!byFile.has(file)) {
+      byFile.set(file, new Set());
+    }
+    byFile.get(file).add(line);
+  };
+  for (const [fileKey, fileEntry] of Object.entries(report)) {
+    if (
+      !isReportFileKey(fileKey) ||
+      !fileEntry ||
+      typeof fileEntry !== "object"
+    ) {
+      continue;
+    }
+    if (fileEntry.scope === "test" && !includeTests) {
+      continue;
+    }
+    const platform = fileEntry.platform || (fileEntry.platforms || [])[0];
+    const allowed = allowedPurlsFor(fileEntry, allowedPurlsByModule);
+    const scope = fileEntry.scope || "main";
+    for (const reference of fileEntry.references || []) {
+      addOccurrence(
+        joinIndex.lookup(reference.symbol, platform, allowed),
+        fileKey,
+        reference.line,
+        "reference",
+        scope,
+        platform,
+      );
+    }
+    for (const call of fileEntry.calls || []) {
+      addOccurrence(
+        joinIndex.lookup(call.owner, platform, allowed),
+        fileKey,
+        call.line,
+        "call",
+        scope,
+        platform,
+      );
+    }
+  }
+  for (const [purl, byFile] of linesByPurl) {
+    const locations = (purlLocationMap[purl] ??= new Set());
+    const files = [...byFile.entries()].map(([file, lines]) => [
+      file,
+      [...lines].sort((left, right) => left - right),
+    ]);
+    // Every file first, then the further lines of each file in turn.
+    for (let depth = 0; locations.size < cap; depth++) {
+      let added = false;
+      for (const [file, lines] of files) {
+        if (depth < lines.length && locations.size < cap) {
+          locations.add(`${file}#${lines[depth]}`);
+          added = true;
+        }
+      }
+      if (!added) {
+        break;
+      }
+    }
+  }
+  for (const purl of Object.keys(scopesByPurl)) {
+    addPropertyValue(
+      componentPropertiesMap,
+      purl,
+      "cdx:scalasem:usageScopes",
+      sortedCsv(scopesByPurl[purl]),
+    );
+    addPropertyValue(
+      componentPropertiesMap,
+      purl,
+      "cdx:scalasem:platforms",
+      sortedCsv(platformsByPurl[purl]),
+    );
+    if (callSitesByPurl[purl]) {
+      addPropertyValue(
+        componentPropertiesMap,
+        purl,
+        "cdx:scalasem:callSites",
+        callSitesByPurl[purl],
+      );
+    }
+  }
+}
+
+/**
+ * Call stacks: each stack the report found becomes one frame list on the
+ * components of the library it ends in, with the entry point first and the
+ * library call last. CycloneDX holds a single call stack per component, and
+ * `framePicker` chooses which of the stacks a component collected to publish.
+ * A library called on the sink's line is reached by the stack too, but a
+ * stack that ends in a call to the component itself is the better evidence:
+ * such a component is offered only its own stacks. Stacks that differ only in
+ * which call of one method into the library they end at are one path; the
+ * method's last call stands for it.
+ */
+function addCallStackEvidence(
+  report,
+  joinIndex,
+  dataFlowFrames,
+  allowedPurlsByModule,
+  options = {},
+) {
+  const includeTests = options.scalasemIncludeTests === true;
+  const ownStacks = {};
+  const reachedStacks = {};
+  for (const stack of report.callStacks || []) {
+    const sink = stack.sink;
+    if (!sink?.owner) {
+      continue;
+    }
+    const fileEntry = report[sink.file];
+    if (fileEntry?.scope === "test" && !includeTests) {
+      continue;
+    }
+    const platform = fileEntry?.platform || (fileEntry?.platforms || [])[0];
+    const allowed = allowedPurlsFor(fileEntry, allowedPurlsByModule);
+    const sinkPurls = new Set(
+      joinIndex.lookupAll(sink.owner, platform, allowed),
+    );
+    const purls = stackTargetPurls(
+      sink,
+      fileEntry,
+      joinIndex,
+      platform,
+      allowed,
+    );
+    if (!purls.length) {
+      continue;
+    }
+    const frames = (stack.frames || [])
+      .filter((frame) => frame?.file)
+      .map((frame) => stackFrame(report, frame));
+    const sinkCovered = frames.some(
+      (frame) => frame.fullFilename === sink.file && frame.line === sink.line,
+    );
+    if (!sinkCovered) {
+      frames.push({
+        package: packageOf(sink.owner),
+        module: sink.owner,
+        function: sink.name || "",
+        line: sink.line,
+        column: undefined,
+        fullFilename: sink.file,
+      });
+    }
+    for (const purl of purls) {
+      const target = sinkPurls.has(purl) ? ownStacks : reachedStacks;
+      target[purl] ??= [];
+      target[purl].push(frames);
+    }
+  }
+  for (const purl of new Set([
+    ...Object.keys(ownStacks),
+    ...Object.keys(reachedStacks),
+  ])) {
+    dataFlowFrames[purl] = onePerPath(ownStacks[purl] || reachedStacks[purl]);
+  }
+}
+
+function onePerPath(stacks) {
+  const byPath = new Map();
+  for (const frames of stacks) {
+    const last = frames[frames.length - 1];
+    const key = [
+      ...frames
+        .slice(0, -1)
+        .map((frame) => `${frame.fullFilename}#${frame.line}`),
+      `${last.fullFilename}#${last.module}.${last.function}`,
+    ].join(">");
+    const known = byPath.get(key);
+    if (!known || (known[known.length - 1].line || 0) < (last.line || 0)) {
+      byPath.set(key, frames);
+    }
+  }
+  return [...byPath.values()];
+}
+
+/**
+ * One call stack frame in the shape `framePicker` reads. The report names the
+ * method; the definition that encloses the frame's line names its class.
+ */
+function stackFrame(report, frame) {
+  const owner = enclosingOwner(report[frame.file], frame);
+  return {
+    package: owner ? packageOf(owner) : "",
+    module: owner || "",
+    function: frame.function || "",
+    line: frame.line || undefined,
+    column: frame.column || undefined,
+    fullFilename: frame.file,
+  };
+}
+
+function enclosingOwner(fileEntry, frame) {
+  let best;
+  for (const definition of fileEntry?.definitions || []) {
+    if (
+      definition?.kind !== "def" ||
+      definition.name !== frame.function ||
+      !definition.owner ||
+      definition.line > frame.line ||
+      (definition.endLine || definition.line) < frame.line
+    ) {
+      continue;
+    }
+    if (!best || definition.line > best.line) {
+      best = definition;
+    }
+  }
+  return best?.owner;
+}
+
+/**
+ * The components a stack is evidence for: the jars that ship the sink class,
+ * plus the libraries called on the sink's own line, which is the terminal hop
+ * the sink call came through. The Scala runtime is ambient and never a
+ * target.
+ */
+function stackTargetPurls(sink, fileEntry, joinIndex, platform, allowed) {
+  const purls = new Set(joinIndex.lookupAll(sink.owner, platform, allowed));
+  for (const call of fileEntry?.calls || []) {
+    if (call.line !== sink.line || !call.owner) {
+      continue;
+    }
+    for (const purl of joinIndex.lookupAll(call.owner, platform, allowed)) {
+      if (!isStdlibPurl(purl)) {
+        purls.add(purl);
+      }
+    }
+  }
+  return [...purls];
+}
+
+function packageOf(className) {
+  const text = String(className || "").replace(/\$+$/, "");
+  const dot = text.lastIndexOf(".");
+  return dot > 0 ? text.slice(0, dot) : "";
+}
+
+const SCALASEM_CRYPTO_PRIMITIVES = new Set([
+  "drbg",
+  "mac",
+  "block-cipher",
+  "stream-cipher",
+  "signature",
+  "hash",
+  "pke",
+  "xof",
+  "kdf",
+  "key-agree",
+  "kem",
+  "ae",
+  "combiner",
+  "key-wrap",
+  "other",
+  "unknown",
+]);
+
+const SCALASEM_MODES = new Set([
+  "cbc",
+  "ecb",
+  "ccm",
+  "gcm",
+  "cfb",
+  "ofb",
+  "ctr",
+]);
+
+const SCALASEM_PADDING = {
+  nopadding: "raw",
+  pkcs5padding: "pkcs5",
+  pkcs7padding: "pkcs7",
+  pkcs1padding: "pkcs1v15",
+  oaepwithsha1andmgf1: "oaep",
+  ssl3padding: "other",
+  iso10126padding: "other",
+};
+
+// The spellings one purl answers to: sbt components carry the encoded purl and
+// the decoded bom-ref, and dependency refs use either.
+function providerAliases(purl, components = []) {
+  if (!purl) {
+    return [];
+  }
+  const component = components.find((candidate) => candidate.purl === purl);
+  return component?.["bom-ref"] && component["bom-ref"] !== purl
+    ? [component["bom-ref"]]
+    : [];
+}
+
+/**
+ * The component purl of the library a crypto finding names as its provider.
+ * Maven coordinates match on group and base artifact; the short aliases cover
+ * the providers the rules emit for libraries without a coordinate.
+ */
+export function scalasemProviderPurl(provider, components = []) {
+  if (!provider) {
+    return undefined;
+  }
+  let group;
+  let artifactPrefix;
+  if (provider === "bcprov") {
+    group = "org.bouncycastle";
+    artifactPrefix = "bcprov";
+  } else if (provider.includes(":")) {
+    [group, artifactPrefix] = provider.split(":");
+  } else {
+    return undefined;
+  }
+  const base = parseScalaArtifact(artifactPrefix).artifactBase;
+  for (const component of components) {
+    if (!component?.purl) {
+      continue;
+    }
+    const segments = /^pkg:[^/]+\/(?:([^/@?]+)\/)?([^@?]+)/.exec(
+      component.purl,
+    );
+    if (!segments || decodeSafe(segments[1]) !== group) {
+      continue;
+    }
+    const name = parseScalaArtifact(decodeSafe(segments[2])).artifactBase;
+    if (name === base || name.startsWith(`${base}-`)) {
+      return component.purl;
+    }
+  }
+  return undefined;
+}
+
+// Primitives a mode of operation applies to. RSA spells `RSA/ECB/...` in JCA,
+// but a public key cipher has no mode, and calling it ECB would read as the
+// weak block cipher mode.
+const SCALASEM_MODE_PRIMITIVES = new Set(["block-cipher", "ae"]);
+
+function cryptoModeOf(finding) {
+  const mode = String(finding.mode || "").toLowerCase();
+  return mode &&
+    SCALASEM_MODES.has(mode) &&
+    SCALASEM_MODE_PRIMITIVES.has(finding.primitive)
+    ? mode
+    : undefined;
+}
+
+/**
+ * The asset name: the algorithm with its key size, mode and padding, so one
+ * cipher used two ways is two assets, and an AES name that carries its mode
+ * (`AES-GCM`) reads the same as AES with that mode.
+ */
+function cryptoAssetDisplayName(finding) {
+  let base = finding.algorithm;
+  let mode = cryptoModeOf(finding)?.toUpperCase();
+  if (SCALASEM_AES_NAMES.has(base) && base.includes("-")) {
+    [base] = base.split("-");
+    mode ||= finding.algorithm.split("-")[1];
+  }
+  const parts = [base];
+  if (finding.keySize) {
+    parts.push(String(finding.keySize));
+  }
+  if (mode) {
+    parts.push(mode);
+  }
+  const padding = paddingOf(finding.padding);
+  if (finding.primitive === "pke" && padding && padding !== "raw") {
+    parts.push(padding.toUpperCase());
+  }
+  return parts.join("-");
+}
+
+function cryptoAlgorithmProperties(finding) {
+  const properties = {};
+  if (finding.primitive) {
+    const primitive = finding.primitive === "rng" ? "drbg" : finding.primitive;
+    properties.primitive = SCALASEM_CRYPTO_PRIMITIVES.has(primitive)
+      ? primitive
+      : "other";
+  }
+  const mode = cryptoModeOf(finding);
+  if (mode) {
+    properties.mode = mode;
+  }
+  const padding = paddingOf(finding.padding);
+  if (padding) {
+    properties.padding = padding;
+  }
+  if (finding.keySize) {
+    properties.parameterSetIdentifier = String(finding.keySize);
+  }
+  return properties;
+}
+
+function cryptoBomRef(kind, name, identifier) {
+  return `crypto/${kind}/${encodeURIComponent(name)}@${encodeURIComponent(identifier || name)}`;
+}
+
+/**
+ * Crypto findings become CycloneDX crypto assets. Algorithms resolve through
+ * the OID alias table and are skipped without one, protocols and keystores
+ * are their own asset kinds, and everything the registry cannot name becomes
+ * a `cdx:scalasem:crypto:*` property on the providing component instead of an
+ * asset with an invented identifier. A finding the report could not resolve
+ * never becomes an asset.
+ */
+function addCryptoEvidence(
+  report,
+  components,
+  cryptoComponentsByRef,
+  cryptoGeneratePurls,
+  componentPropertiesMap,
+  metadataProperties,
+  options = {},
+) {
+  const includeTests = options.scalasemIncludeTests === true;
+  const nativeLibraryBySymbol = new Map();
+  for (const binding of report.nativeBindings || []) {
+    if (binding.library && binding.symbol) {
+      nativeLibraryBySymbol.set(binding.symbol, binding.library);
+    }
+  }
+  const assetOccurrencesByRef = new Map();
+  for (const finding of report.crypto || []) {
+    const scope = report[finding.file]?.scope || "main";
+    if (scope === "test" && !includeTests) {
+      continue;
+    }
+    const location = `${finding.file || ""}#${finding.line || 0}`;
+    const nativeLibrary = finding.kind?.startsWith("native")
+      ? nativeLibraryBySymbol.get(finding.name)
+      : undefined;
+    const providerPurl = scalasemProviderPurl(finding.provider, components);
+    const asset =
+      finding.resolution === "unresolved"
+        ? undefined
+        : cryptoAssetFor(finding, options);
+    if (asset) {
+      const existing = cryptoComponentsByRef.get(asset["bom-ref"]);
+      const target = existing || asset;
+      if (!existing) {
+        cryptoComponentsByRef.set(asset["bom-ref"], asset);
+      }
+      for (const property of cryptoAssetProperties(finding, nativeLibrary)) {
+        appendUniqueProperty(target.properties, property.name, property.value);
+      }
+      appendUniqueProperty(
+        target.properties,
+        "cdx:scalasem:usageScopes",
+        scope,
+      );
+      const occurrences = assetOccurrencesByRef.get(asset["bom-ref"]) || [];
+      if (!occurrences.includes(location)) {
+        occurrences.push(location);
+      }
+      assetOccurrencesByRef.set(asset["bom-ref"], occurrences);
+      if (providerPurl) {
+        cryptoGeneratePurls[providerPurl] ??= new Set();
+        cryptoGeneratePurls[providerPurl].add(asset["bom-ref"]);
+        addPropertyValue(
+          componentPropertiesMap,
+          providerPurl,
+          "cdx:scalasem:cryptoAlgorithms",
+          finding.algorithm,
+        );
+      }
+      providerAliases(providerPurl, components).forEach((alias) => {
+        cryptoGeneratePurls[alias] ??= new Set();
+        cryptoGeneratePurls[alias].add(asset["bom-ref"]);
+      });
+    } else if (finding.algorithm || finding.name) {
+      // No registry OID: the finding stays visible as a property, and a weak
+      // one says so, since no asset carries the verdict.
+      const name = finding.algorithm || finding.name;
+      const value = `${name}@${location}`;
+      const record = (propertyName) => {
+        if (providerPurl) {
+          addPropertyValue(
+            componentPropertiesMap,
+            providerPurl,
+            propertyName,
+            value,
+          );
+        } else {
+          appendUniqueProperty(metadataProperties, propertyName, value);
+        }
+      };
+      record(`cdx:scalasem:crypto:${finding.kind}`);
+      if (finding.weak) {
+        record("cdx:scalasem:crypto:weakFinding");
+      }
+    }
+  }
+  for (const [ref, asset] of cryptoComponentsByRef.entries()) {
+    asset.evidence = {
+      occurrences: assetOccurrencesByRef
+        .get(ref)
+        .sort()
+        .map((location) => parseLocationString(location)),
+    };
+  }
+}
+
+function parseLocationString(location) {
+  const separator = location.lastIndexOf("#");
+  const line = Number.parseInt(location.slice(separator + 1), 10);
+  return {
+    location: location.slice(0, separator),
+    ...(Number.isFinite(line) && line > 0 ? { line } : {}),
+  };
+}
+
+// Protocol names as CycloneDX types and versions. SSL predates TLS and has
+// no type of its own, and the bare names carry no version.
+const SCALASEM_PROTOCOLS = {
+  SSL: { type: "other" },
+  SSLv3: { type: "other", version: "3.0" },
+  TLS: { type: "tls" },
+  TLSv1: { type: "tls", version: "1.0" },
+  "TLSv1.1": { type: "tls", version: "1.1" },
+  "TLSv1.2": { type: "tls", version: "1.2" },
+  "TLSv1.3": { type: "tls", version: "1.3" },
+  "DTLSv1.2": { type: "dtls", version: "1.2" },
+};
+
+function cryptoAssetFor(finding) {
+  const name = finding.algorithm || finding.name;
+  if (!name) {
+    return undefined;
+  }
+  if (SCALASEM_PROTOCOL_NAMES.has(name) || finding.kind === "protocol") {
+    const protocol = SCALASEM_PROTOCOLS[name] || { type: "other" };
+    return {
+      type: "cryptographic-asset",
+      name,
+      "bom-ref": cryptoBomRef("protocol", name),
+      description:
+        "Cryptographic protocol detected by scalasem source analysis",
+      cryptoProperties: {
+        assetType: "protocol",
+        protocolProperties: { ...protocol },
+      },
+      properties: [],
+    };
+  }
+  if (SCALASEM_KEYSTORE_NAMES.has(name) || finding.kind === "keystore") {
+    return {
+      type: "cryptographic-asset",
+      name,
+      "bom-ref": cryptoBomRef("keystore", name),
+      description: "Key store detected by scalasem source analysis",
+      cryptoProperties: {
+        assetType: "related-crypto-material",
+        relatedCryptoMaterialProperties: { type: "other" },
+      },
+      properties: [],
+    };
+  }
+  const oidKey = scalasemCryptoOidKey(
+    name,
+    finding.keySize,
+    finding.mode,
+    finding.curve,
+    finding.padding,
+  );
+  const oid = oidKey ? SCALASEM_CRYPTO_OIDS[oidKey]?.oid : undefined;
+  if (!oid) {
+    return undefined;
+  }
+  const algorithmProperties = cryptoAlgorithmProperties(finding);
+  const displayName = cryptoAssetDisplayName(finding);
+  // A native finding's bom-ref carries the C function it resolves to, so the
+  // asset stays distinct per binding and findable by the function name.
+  const nativeSuffix =
+    finding.kind?.startsWith("native") && finding.name
+      ? `#${finding.name}`
+      : "";
+  const component = {
+    type: "cryptographic-asset",
+    name: displayName,
+    "bom-ref": `${cryptoBomRef("algorithm", displayName, oid)}${nativeSuffix}`,
+    description: "Cryptographic algorithm detected by scalasem source analysis",
+    cryptoProperties: {
+      assetType: "algorithm",
+      oid,
+      algorithmProperties,
+    },
+    properties: [],
+  };
+  // The family and the curve in the schema's own spellings (`secg/secp256r1`);
+  // spec compatibility keeps them for 1.7 and later only.
+  applyAlgorithmProperties(component, {
+    name: SCALASEM_AES_NAMES.has(name) ? "AES" : name,
+    curve: finding.curve,
+  });
+  return component;
+}
+
+function cryptoAssetProperties(finding, nativeLibrary) {
+  const properties = [];
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:crypto:resolution",
+    finding.resolution,
+  );
+  appendUniqueProperty(properties, "cdx:scalasem:crypto:api", finding.api);
+  appendUniqueProperty(
+    properties,
+    "cdx:scalasem:crypto:provider",
+    finding.provider,
+  );
+  if (finding.weak) {
+    appendUniqueProperty(properties, "cdx:scalasem:crypto:weak", "true");
+  }
+  if (finding.name && finding.kind?.startsWith("native")) {
+    appendUniqueProperty(
+      properties,
+      "cdx:scalasem:crypto:nativeFunction",
+      finding.name,
+    );
+    appendUniqueProperty(
+      properties,
+      "cdx:scalasem:crypto:nativeLibrary",
+      nativeLibrary,
+    );
+  }
+  if (finding.bits) {
+    appendUniqueProperty(
+      properties,
+      "cdx:scalasem:crypto:gcmTagBits",
+      finding.bits,
+    );
+  }
+  return properties;
+}
+
+/**
+ * A JDBC address without what may carry credentials: userinfo, the Oracle
+ * `user/password@` prefix, URL parameters and the `;key=value` properties
+ * some drivers take. The report is sanitized already; this keeps the BOM
+ * safe from a report written by anything else.
+ *
+ * @param {string} value JDBC URL.
+ * @returns {string} The address part.
+ */
+export function sanitizeJdbcUrl(value) {
+  let text = String(value || "").trim();
+  if (!/^jdbc:/i.test(text)) {
+    return text;
+  }
+  text = `jdbc:${text.slice(5)}`;
+  for (const separator of ["?", "#", ";"]) {
+    const index = text.indexOf(separator);
+    if (index >= 0) {
+      text = text.slice(0, index);
+    }
+  }
+  const slashes = text.indexOf("//");
+  if (slashes >= 0) {
+    const authorityEnd = text.indexOf("/", slashes + 2);
+    const authority = text.slice(
+      slashes + 2,
+      authorityEnd >= 0 ? authorityEnd : undefined,
+    );
+    const at = authority.lastIndexOf("@");
+    if (at >= 0) {
+      text =
+        text.slice(0, slashes + 2) +
+        authority.slice(at + 1) +
+        (authorityEnd >= 0 ? text.slice(authorityEnd) : "");
+    }
+  }
+  // Oracle thin: jdbc:oracle:thin:user/password@host:port:sid
+  const thin = /^(jdbc:oracle:thin:)[^@]*@/i.exec(text);
+  if (thin) {
+    text = `${thin[1]}@${text.slice(thin[0].length)}`;
+  }
+  return text;
+}
+
+function jdbcServiceName(value) {
+  const address = sanitizeJdbcUrl(value);
+  const scheme = address.slice(5).split(":", 1)[0].toLowerCase() || "jdbc";
+  let host = /\/\/([^/:]+)/.exec(address)?.[1];
+  if (!host) {
+    // Oracle thin and similar: the host follows the @.
+    host = /@(?:\/\/)?([^/:]+)/.exec(address)?.[1];
+  }
+  return [scheme, host].filter(Boolean).join("-");
+}
+
+/**
+ * The host a service URL names, or the topic and JDBC scheme equivalents, so
+ * every outbound service is named after where it talks to rather than the
+ * client library it uses.
+ */
+export function scalasemServiceName(service) {
+  if (service?.topic) {
+    return scalasemServiceEndpoint(service);
+  }
+  const value = service?.url || service?.host;
+  if (!value) {
+    return undefined;
+  }
+  if (/^jdbc:/i.test(value)) {
+    return jdbcServiceName(value);
+  }
+  if (service?.host) {
+    return service.host.split(":")[0];
+  }
+  try {
+    return new URL(value).hostname || undefined;
+  } catch (_err) {
+    return undefined;
+  }
+}
+
+/**
+ * The endpoint a service row carries, normalized: a URL as its scheme and
+ * host, a host as it is, a topic as `client:topic`, a JDBC URL with its
+ * credentials and parameters dropped.
+ */
+export function scalasemServiceEndpoint(service) {
+  if (service?.topic) {
+    const scheme = service.client || "messaging";
+    return `${scheme}:${service.topic}`;
+  }
+  const value = service?.url || service?.host;
+  if (!value) {
+    return undefined;
+  }
+  if (/^jdbc:/i.test(value)) {
+    // The database name in the path is part of the address.
+    return sanitizeJdbcUrl(value);
+  }
+  if (service?.host) {
+    return service.host;
+  }
+  try {
+    const parsed = new URL(value);
+    return `${parsed.protocol}//${parsed.host}`;
+  } catch (_err) {
+    const hostMatch = /:\/\/([^/?#@]+@)?([^/?#]+)/.exec(value);
+    return hostMatch ? `${value.split("://")[0]}://${hostMatch[2]}` : undefined;
+  }
+}
+
+function serviceLocation(service) {
+  if (!service?.file) {
+    return undefined;
+  }
+  return service.line ? `${service.file}#${service.line}` : service.file;
+}
+
+/**
+ * Outbound services: HTTP and websocket clients, data stores, messaging
+ * topics and cloud clients. Locations ride in
+ * `cdx:scalasem:service:location` properties because CycloneDX services
+ * carry no evidence field before 2.0; spec compatibility keeps the
+ * occurrences below that only at 2.0.
+ *
+ * @param {Object} report Parsed scalasem report.
+ * @param {Object} servicesMap Map populated with service definitions.
+ * @returns {Object} The mutated services map.
+ */
+export function collectScalasemServices(report = {}, servicesMap = {}) {
+  for (const service of report.services || []) {
+    const name = scalasemServiceName(service);
+    if (!name) {
+      continue;
+    }
+    const definition = (servicesMap[name] ??= {
+      name,
+      endpoints: new Set(),
+      authenticated: undefined,
+      xTrustBoundary: true,
+      properties: [],
+    });
+    definition.endpoints ??= new Set();
+    definition.properties ??= [];
+    const value = scalasemServiceEndpoint(service);
+    if (value) {
+      definition.endpoints.add(value);
+    }
+    const location = serviceLocation(service);
+    if (location) {
+      appendUniqueProperty(
+        definition.properties,
+        "cdx:scalasem:service:location",
+        location,
+      );
+      definition.evidence ??= { occurrences: [] };
+      const known = definition.evidence.occurrences;
+      if (!known.some((occurrence) => occurrence.location === location)) {
+        known.push(parseLocationString(location));
+      }
+    }
+    appendUniqueProperty(
+      definition.properties,
+      "cdx:scalasem:service:kind",
+      service.kind,
+    );
+    appendUniqueProperty(
+      definition.properties,
+      "cdx:scalasem:service:client",
+      service.client,
+    );
+    appendUniqueProperty(
+      definition.properties,
+      "cdx:scalasem:service:resolution",
+      service.resolution,
+    );
+  }
+  return servicesMap;
+}
+
+/**
+ * Inbound endpoints, one service per route, named the way the OpenAPI reader
+ * names its own so the two converge on one entry for the same route.
+ *
+ * @param {Object} report Parsed scalasem report.
+ * @param {Object} servicesMap Map populated with service definitions.
+ * @returns {Object} The mutated services map.
+ */
+export function collectScalasemApiEndpoints(report = {}, servicesMap = {}) {
+  for (const endpoint of report.endpoints || []) {
+    const path = endpoint?.path;
+    const method = String(endpoint.method || "").toUpperCase();
+    if (!path || !method) {
+      continue;
+    }
+    const serviceName = `service-${path.replaceAll("/", "")}-${method.toLowerCase()}`;
+    const definition = (servicesMap[serviceName] ??= {
+      name: serviceName,
+      endpoints: new Set(),
+      authenticated: undefined,
+      xTrustBoundary: undefined,
+      properties: [],
+    });
+    definition.endpoints ??= new Set();
+    definition.properties ??= [];
+    definition.endpoints.add(path);
+    if (endpoint.authenticated === true) {
+      definition.authenticated = true;
+      definition.xTrustBoundary = true;
+    }
+    if (endpoint.handler) {
+      appendUniqueProperty(
+        definition.properties,
+        "internal:operationId",
+        endpoint.handler,
+      );
+      appendUniqueProperty(
+        definition.properties,
+        "cdx:scalasem:endpoint:handler",
+        endpoint.handler,
+      );
+    }
+    appendUniqueProperty(
+      definition.properties,
+      "cdx:service:httpMethod",
+      method,
+    );
+    appendUniqueProperty(
+      definition.properties,
+      "cdx:scalasem:endpoint:framework",
+      endpoint.framework,
+    );
+    const location = serviceLocation(endpoint);
+    if (location) {
+      appendUniqueProperty(
+        definition.properties,
+        "cdx:scalasem:service:location",
+        location,
+      );
+      definition.evidence ??= { occurrences: [] };
+      const known = definition.evidence.occurrences;
+      if (!known.some((occurrence) => occurrence.location === location)) {
+        known.push(parseLocationString(location));
+      }
+    }
+  }
+  return servicesMap;
+}
+
+/**
+ * The npm packages a Scala.js build bundles. scalajs-bundler installs them
+ * with a lock file under `target`, and a bundler workspace such as a Vite
+ * client keeps its own manifest and lock file beside the build. Their npm
+ * components join the BOM so the report's JavaScript module imports can be
+ * attributed to them. Only lock files name the installed versions, and a
+ * workspace counts only when its manifest uses Scala.js, so a documentation
+ * site or a tool's own `package.json` adds nothing.
+ *
+ * @param {string} src Project directory.
+ * @param {Object[]} pkgList Components the build tools reported.
+ * @param {Object} [options] CLI options; `exclude` is honoured.
+ * @returns {Promise<{components: Object[], roots: string[], dependencies: Object[]}>}
+ *   Npm components, the bom-refs the workspaces depend on directly, and the
+ *   lock files' dependency edges between the npm components.
+ */
+export async function collectScalaJsNpmComponents(
+  src,
+  pkgList = [],
+  options = {},
+) {
+  const isScalaJs = pkgList.some((pkg) => {
+    const name = pkg?.name || purlPurlName(pkg?.purl) || "";
+    return /_(sjs\d+(?:\.\d+)?)$/.test(name) || name === "scalajs-library";
+  });
+  const empty = { components: [], roots: [], dependencies: [] };
+  if (!isScalaJs) {
+    return empty;
+  }
+  const projectDir = resolve(src);
+  const searchOptions = {
+    includeNodeModulesDir: false,
+    exclude: options.exclude,
+  };
+  const manifestDirs = new Set();
+  for (const file of getAllFiles(
+    projectDir,
+    "**/scalajs-bundler/**/package.json",
+    searchOptions,
+  )) {
+    manifestDirs.add(dirname(file));
+  }
+  // A workspace beside the build: the project's own manifests, not the ones
+  // node_modules holds. Two levels cover the usual client/server split.
+  for (const pattern of ["package.json", "*/package.json"]) {
+    for (const file of getAllFiles(projectDir, pattern, searchOptions)) {
+      if (manifestUsesScalaJs(file)) {
+        manifestDirs.add(dirname(file));
+      }
+    }
+  }
+  const components = [];
+  const roots = [];
+  const dependencies = [];
+  for (const manifestDir of manifestDirs) {
+    const lockFile = ["package-lock.json", "npm-shrinkwrap.json"]
+      .map((name) => join(manifestDir, name))
+      .find((file) => safeExistsSync(file));
+    if (!lockFile) {
+      continue;
+    }
+    const lockData = await parsePkgLock(lockFile, options);
+    // The lock file's own package is the workspace, which is part of the
+    // project rather than a dependency of it.
+    const workspaceRefs = new Set(
+      (lockData?.pkgList || [])
+        .filter((pkg) => pkg.type === "application")
+        .map((pkg) => pkg["bom-ref"]),
+    );
+    components.push(
+      ...(lockData?.pkgList || []).filter((pkg) => pkg.type !== "application"),
+    );
+    for (const dependency of lockData?.dependenciesList || []) {
+      if (workspaceRefs.has(dependency.ref)) {
+        roots.push(...(dependency.dependsOn || []));
+      } else {
+        dependencies.push(dependency);
+      }
+    }
+  }
+  return { components, roots: [...new Set(roots)], dependencies };
+}
+
+function manifestUsesScalaJs(manifestFile) {
+  try {
+    const manifest = JSON.parse(readFileSync(manifestFile, "utf-8"));
+    const names = Object.keys({
+      ...(manifest?.dependencies || {}),
+      ...(manifest?.devDependencies || {}),
+    });
+    return names.some((name) => /scala-?js/i.test(name));
+  } catch (_err) {
+    return false;
+  }
+}
+
+// The package name an npm purl carries, with the scope decoded back to
+// `@scope/name`, which is how JavaScript names its modules.
+function npmNameFromPurl(purl) {
+  const rest = /^pkg:npm\/(.+)$/.exec(purl || "")?.[1];
+  if (!rest) {
+    return undefined;
+  }
+  const at = rest.lastIndexOf("@");
+  const name = at > 0 ? rest.slice(0, at) : rest;
+  try {
+    return decodeURIComponent(name);
+  } catch (_err) {
+    return name;
+  }
+}
+
+/**
+ * Attribute the report's JavaScript module imports to the npm components the
+ * build bundles. A module no component owns is reported on the project
+ * component instead, so the import stays visible either way.
+ */
+function addJsModuleEvidence(
+  report,
+  components,
+  purlLocationMap,
+  componentPropertiesMap,
+  metadataProperties,
+) {
+  const jsModules = report.jsModules || [];
+  if (!jsModules.length) {
+    return;
+  }
+  const npmByName = new Map();
+  for (const component of components) {
+    const name = npmNameFromPurl(component?.purl);
+    if (name) {
+      npmByName.set(name, component);
+    }
+  }
+  const moduleMatchesComponent = (moduleName) => {
+    if (npmByName.has(moduleName)) {
+      return npmByName.get(moduleName);
+    }
+    // `pkg/subpath` and `@scope/pkg/subpath` belong to the package named by
+    // their leading segments.
+    const segments = moduleName.split("/");
+    const scoped =
+      moduleName.startsWith("@") && segments.length > 2
+        ? `${segments[0]}/${segments[1]}`
+        : segments[0];
+    return npmByName.get(scoped);
+  };
+  const unmatched = new Set();
+  const seenLocations = new Set();
+  for (const jsModule of jsModules) {
+    const moduleName = jsModule.module;
+    if (!moduleName || moduleName.startsWith(".") || moduleName.includes("*")) {
+      continue;
+    }
+    const component = moduleMatchesComponent(moduleName);
+    const location = jsModule.line
+      ? `${jsModule.file}#${jsModule.line}`
+      : jsModule.file;
+    if (!component) {
+      unmatched.add(moduleName);
+      continue;
+    }
+    if (!seenLocations.has(`${component.purl}#${location}`)) {
+      seenLocations.add(`${component.purl}#${location}`);
+      addSetValue(purlLocationMap, component.purl, location);
+    }
+    addPropertyValue(
+      componentPropertiesMap,
+      component.purl,
+      "cdx:scalasem:jsModule",
+      moduleName,
+    );
+  }
+  if (unmatched.size) {
+    appendUniqueProperty(
+      metadataProperties,
+      "cdx:scalasem:jsModules",
+      sortedCsv(unmatched),
+    );
+  }
+}