fix: IaC path parsing

Fix a bug in which IaC resource paths (more context below) was naively
split by dot, without distinguishing between separator dots and dots
inside string literals.

For example: "foo.bar.baz" was split into 3 words correctly, but
"metadata.annotations['container.apparmor.security.beta.kubernetes.io/web']"
was not.

The "path" / "cloud config path" is a sort of address used to identify
locations in files where vulnerabilities reside. It is very similar to a
jsonpath. The user receives it as an array of components, which in the
future can be useful for building tooling around, e.g. the upcoming CLI
ignores.

The code that evaluates policies against source files is also
responsible for returning a representation of this path, which it does
as a dot-separated string.

package.json

@@ -114,6 +114,7 @@
     "open": "^7.0.3",
     "ora": "5.4.0",
     "os-name": "^3.0.0",
+    "pegjs": "^0.10.0",
     "promise-queue": "^2.2.5",
     "proxy-from-env": "^1.0.0",
     "rimraf": "^2.6.3",

src/cli/commands/test/iac-local-execution/parsers/path.ts

@@ -0,0 +1,35 @@
+import * as peg from 'pegjs';
+
+const grammar = `
+start
+  = element+
+
+element
+  = component:component "."? { return component; }
+
+component
+  = $(identifier index?)
+
+identifier
+  = $([^'"\\[\\]\\.]+)
+
+index
+  = $("[" ['"]? [^'"\\]]+ ['"]? "]")
+`;
+
+export const parsePath = createPathParser();
+
+function createPathParser(): (expr: string) => string[] {
+  const parser = peg.generate(grammar);
+  return (expr: string) => {
+    try {
+      return parser.parse(expr);
+    } catch (e) {
+      // I haven't actually been able to write a testcase that triggers this
+      // code path, but I've included it anyway as a fallback to allow users to
+      // keep using the CLI even if this does occur. Their paths might look
+      // strange, but that's better than nothing.
+      return expr.split('.');
+    }
+  };
+}

src/cli/commands/test/iac-local-execution/results-formatter.ts

@@ -7,6 +7,7 @@ import {
   PolicyMetadata,
   TestMeta,
 } from './types';
+import { parsePath } from './parsers/path';
 import * as path from 'path';
 import { SEVERITY } from '../../../../lib/snyk-test/common';
 import { IacProjectType } from '../../../../lib/iac/constants';
@@ -55,7 +56,7 @@ function formatScanResult(
   const formattedIssues = scanResult.violatedPolicies.map((policy) => {
     const cloudConfigPath =
       scanResult.docId !== undefined
-        ? [`[DocId: ${scanResult.docId}]`].concat(policy.msg.split('.'))
+        ? [`[DocId: ${scanResult.docId}]`].concat(parsePath(policy.msg))
         : policy.msg.split('.');
 
     const flagsRequiringLineNumber = [

test/jest/unit/iac-unit-tests/path-parser.spec.ts

@@ -0,0 +1,19 @@
+import { parsePath } from '../../../../src/cli/commands/test/iac-local-execution/parsers/path';
+
+describe('parsing cloudConfigPath', () => {
+  it.each([
+    ['foo', ['foo']],
+    ['foo.bar.baz', ['foo', 'bar', 'baz']],
+    ['foo_1._bar2.baz3_', ['foo_1', '_bar2', 'baz3_']],
+    ['foo.bar[abc].baz', ['foo', 'bar[abc]', 'baz']],
+    ['foo.bar[abc.def].baz', ['foo', 'bar[abc.def]', 'baz']],
+    ["foo.bar['abc.def'].baz", ['foo', "bar['abc.def']", 'baz']],
+    ['foo.bar["abc.def"].baz', ['foo', 'bar["abc.def"]', 'baz']],
+    ["foo.bar['abc/def'].baz", ['foo', "bar['abc/def']", 'baz']],
+    ["foo.bar['abcdef'].baz", ['foo', "bar['abcdef']", 'baz']],
+    ["bar['abc.def']", ["bar['abc.def']"]],
+    ["fo%o.bar['ab$c/def'].baz", ['fo%o', "bar['ab$c/def']", 'baz']],
+  ])('%s', (input, expected) => {
+    expect(parsePath(input)).toEqual(expected);
+  });
+});