[Security Solution] Fixes multi-line diff algorithm performance in th…

…e `upgrade/_review` endpoint (elastic#199388) **Fixes elastic#199290 ## Summary The current multi-line string algorithm uses a very inefficient regex to split and analyze string fields, and exponentially increases in time complexity when the strings are long. This PR substitutes a much simpler comparison regex for far better efficiency as shown in the table below. ### Performance between different regex options using sample prebuilt rule setup guide string | | `/(\S+\|\s+)/g` (original) | `/(\s+)/g` | `/(\n)/g` | `/(\r\n\|\n\|\r)/g` | |-----------------------|---------------|----------|---------|-------------------| | Unit test speed | `986ms` | `96ms` | `1ms` | `2ms` | | FTR test with 1 rule | `3.0s` | `2.8s` | `2.0s` | `2.0s` | | FTR test with 5 rules | `11.6s` | `6.8s` | `6.1s` | | ### Performance between different regex options using intentionally long strings (25k characters) | | `/(\S+\|\s+)/g` | `/(\r\n\|\n\|\r)/g` | |----------------------|-----------------------|---------------------| | Unit test speed | `1049414ms` (17 min) | `58ms` | | FTR test with 1 rule | `>360000ms` (Timeout) | `2.1 s` | ### Checklist Delete any items that are not applicable to this PR. - [x] [Unit or functional tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html) were updated or added to match the most common scenarios - [ ] [Flaky Test Runner](https://ci-stats.kibana.dev/trigger_flaky_test_runner/1) was used on any tests changed ### For maintainers - [ ] This was checked for breaking API changes and was [labeled appropriately](https://www.elastic.co/guide/en/kibana/master/contributing.html#_add_your_labels) - [ ] This will appear in the **Release Notes** and follow the [guidelines](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process) --------- Co-authored-by: Elastic Machine <elasticmachine@users.noreply.github.com> Co-authored-by: Georgii Gorbachev <georgii.gorbachev@elastic.co>
kibanamachine · Nov 13, 2024 · 4f6d357 · 4f6d357
1 parent c35b389
commit 4f6d357
Show file tree

Hide file tree

Showing 4 changed files with 202 additions and 26 deletions.
diff --git a/...prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.mock.ts b/...prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.mock.ts
diff --git a/...prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.test.ts b/...prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.test.ts
@@ -13,13 +13,23 @@ import {
   ThreeWayDiffConflict,
 } from '../../../../../../../../common/api/detection_engine';
 import { multiLineStringDiffAlgorithm } from './multi_line_string_diff_algorithm';
+import {
+  TEXT_M_A,
+  TEXT_M_B,
+  TEXT_M_C,
+  TEXT_M_MERGED,
+  TEXT_XL_A,
+  TEXT_XL_B,
+  TEXT_XL_C,
+  TEXT_XL_MERGED,
+} from './multi_line_string_diff_algorithm.mock';
 
 describe('multiLineStringDiffAlgorithm', () => {
   it('returns current_version as merged output if there is no update - scenario AAA', () => {
     const mockVersions: ThreeVersionsOf<string> = {
-      base_version: 'My description.\nThis is a second line.',
-      current_version: 'My description.\nThis is a second line.',
-      target_version: 'My description.\nThis is a second line.',
+      base_version: TEXT_M_A,
+      current_version: TEXT_M_A,
+      target_version: TEXT_M_A,
     };
 
     const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -36,9 +46,9 @@ describe('multiLineStringDiffAlgorithm', () => {
 
   it('returns current_version as merged output if current_version is different and there is no update - scenario ABA', () => {
     const mockVersions: ThreeVersionsOf<string> = {
-      base_version: 'My description.\nThis is a second line.',
-      current_version: 'My GREAT description.\nThis is a second line.',
-      target_version: 'My description.\nThis is a second line.',
+      base_version: TEXT_M_A,
+      current_version: TEXT_M_B,
+      target_version: TEXT_M_A,
     };
 
     const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -55,9 +65,9 @@ describe('multiLineStringDiffAlgorithm', () => {
 
   it('returns target_version as merged output if current_version is the same and there is an update - scenario AAB', () => {
     const mockVersions: ThreeVersionsOf<string> = {
-      base_version: 'My description.\nThis is a second line.',
-      current_version: 'My description.\nThis is a second line.',
-      target_version: 'My GREAT description.\nThis is a second line.',
+      base_version: TEXT_M_A,
+      current_version: TEXT_M_A,
+      target_version: TEXT_M_B,
     };
 
     const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -74,9 +84,9 @@ describe('multiLineStringDiffAlgorithm', () => {
 
   it('returns current_version as merged output if current version is different but it matches the update - scenario ABB', () => {
     const mockVersions: ThreeVersionsOf<string> = {
-      base_version: 'My description.\nThis is a second line.',
-      current_version: 'My GREAT description.\nThis is a second line.',
-      target_version: 'My GREAT description.\nThis is a second line.',
+      base_version: TEXT_M_A,
+      current_version: TEXT_M_B,
+      target_version: TEXT_M_B,
     };
 
     const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -92,32 +102,53 @@ describe('multiLineStringDiffAlgorithm', () => {
   });
 
   describe('if all three versions are different - scenario ABC', () => {
-    it('returns a computated merged version without a conflict if 3 way merge is possible', () => {
+    it('returns a computated merged version with a solvable conflict if 3 way merge is possible (real-world example)', () => {
       const mockVersions: ThreeVersionsOf<string> = {
-        base_version: `My description.\f\nThis is a second\u2001 line.\f\nThis is a third line.`,
-        current_version: `My GREAT description.\f\nThis is a second\u2001 line.\f\nThis is a third line.`,
-        target_version: `My description.\f\nThis is a second\u2001 line.\f\nThis is a GREAT line.`,
+        base_version: TEXT_M_A,
+        current_version: TEXT_M_B,
+        target_version: TEXT_M_C,
       };
 
-      const expectedMergedVersion = `My GREAT description.\f\nThis is a second\u2001 line.\f\nThis is a GREAT line.`;
+      const result = multiLineStringDiffAlgorithm(mockVersions);
+
+      expect(result).toEqual(
+        expect.objectContaining({
+          merged_version: TEXT_M_MERGED,
+          diff_outcome: ThreeWayDiffOutcome.CustomizedValueCanUpdate,
+          conflict: ThreeWayDiffConflict.SOLVABLE,
+          merge_outcome: ThreeWayMergeOutcome.Merged,
+        })
+      );
+    });
+
+    it('returns a computated merged version with a solvable conflict if 3 way merge is possible (simplified example)', () => {
+      // 3 way merge is possible when changes are made to different lines of text
+      // (in other words, there are no different changes made to the same line of text).
+      const mockVersions: ThreeVersionsOf<string> = {
+        base_version: 'My description.\nThis is a second line.',
+        current_version: 'My MODIFIED description.\nThis is a second line.',
+        target_version: 'My description.\nThis is a MODIFIED second line.',
+      };
 
       const result = multiLineStringDiffAlgorithm(mockVersions);
 
       expect(result).toEqual(
         expect.objectContaining({
-          merged_version: expectedMergedVersion,
+          merged_version: 'My MODIFIED description.\nThis is a MODIFIED second line.',
           diff_outcome: ThreeWayDiffOutcome.CustomizedValueCanUpdate,
           conflict: ThreeWayDiffConflict.SOLVABLE,
           merge_outcome: ThreeWayMergeOutcome.Merged,
         })
       );
     });
 
-    it('returns the current_version with a conflict if 3 way merge is not possible', () => {
+    it('returns the current_version with a non-solvable conflict if 3 way merge is not possible (simplified example)', () => {
+      // It's enough to have different changes made to the same line of text
+      // to trigger a NON_SOLVABLE conflict. This behavior is similar to how Git works.
       const mockVersions: ThreeVersionsOf<string> = {
         base_version: 'My description.\nThis is a second line.',
-        current_version: 'My GREAT description.\nThis is a third line.',
-        target_version: 'My EXCELLENT description.\nThis is a fourth.',
+        current_version: 'My GREAT description.\nThis is a second line.',
+        target_version: 'My EXCELLENT description.\nThis is a second line.',
       };
 
       const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -131,14 +162,39 @@ describe('multiLineStringDiffAlgorithm', () => {
         })
       );
     });
+
+    it('does not exceed performance limits when diffing and merging extra large input texts', () => {
+      const mockVersions: ThreeVersionsOf<string> = {
+        base_version: TEXT_XL_A,
+        current_version: TEXT_XL_B,
+        target_version: TEXT_XL_C,
+      };
+
+      const startTime = performance.now();
+      const result = multiLineStringDiffAlgorithm(mockVersions);
+      const endTime = performance.now();
+
+      // If the regex merge in this function takes over 500ms, this test fails
+      // Performance measurements: https://github.com/elastic/kibana/pull/199388
+      expect(endTime - startTime).toBeLessThan(500);
+
+      expect(result).toEqual(
+        expect.objectContaining({
+          merged_version: TEXT_XL_MERGED,
+          diff_outcome: ThreeWayDiffOutcome.CustomizedValueCanUpdate,
+          conflict: ThreeWayDiffConflict.SOLVABLE,
+          merge_outcome: ThreeWayMergeOutcome.Merged,
+        })
+      );
+    });
   });
 
   describe('if base_version is missing', () => {
     it('returns current_version as merged output if current_version and target_version are the same - scenario -AA', () => {
       const mockVersions: ThreeVersionsOf<string> = {
         base_version: MissingVersion,
-        current_version: 'My description.\nThis is a second line.',
-        target_version: 'My description.\nThis is a second line.',
+        current_version: TEXT_M_A,
+        target_version: TEXT_M_A,
       };
 
       const result = multiLineStringDiffAlgorithm(mockVersions);
@@ -158,8 +214,8 @@ describe('multiLineStringDiffAlgorithm', () => {
     it('returns target_version as merged output if current_version and target_version are different - scenario -AB', () => {
       const mockVersions: ThreeVersionsOf<string> = {
         base_version: MissingVersion,
-        current_version: `My GREAT description.\nThis is a second line.`,
-        target_version: `My description.\nThis is a second line, now longer.`,
+        current_version: TEXT_M_A,
+        target_version: TEXT_M_B,
       };
 
       const result = multiLineStringDiffAlgorithm(mockVersions);

diff --git a/...gine/prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.ts b/...gine/prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.ts
@@ -102,7 +102,7 @@ const mergeVersions = ({
       // TS does not realize that in ABC scenario, baseVersion cannot be missing
       // Missing baseVersion scenarios were handled as -AA and -AB.
       const mergedVersion = merge(currentVersion, baseVersion ?? '', targetVersion, {
-        stringSeparator: /(\S+|\s+)/g, // Retains all whitespace, which we keep to preserve formatting
+        stringSeparator: /(\r\n|\n|\r)/g, // Separates strings by new lines
       });
 
       return mergedVersion.conflict

diff --git a/...ent/trial_license_complete_tier/upgrade_review_prebuilt_rules.multi_line_string_fields.ts b/...ent/trial_license_complete_tier/upgrade_review_prebuilt_rules.multi_line_string_fields.ts
@@ -10,6 +10,12 @@ import {
   ThreeWayDiffOutcome,
   ThreeWayMergeOutcome,
 } from '@kbn/security-solution-plugin/common/api/detection_engine';
+import {
+  TEXT_XL_A,
+  TEXT_XL_B,
+  TEXT_XL_C,
+  TEXT_XL_MERGED,
+} from '@kbn/security-solution-plugin/server/lib/detection_engine/prebuilt_rules/logic/diff/calculation/algorithms/multi_line_string_diff_algorithm.mock';
 import { FtrProviderContext } from '../../../../../../ftr_provider_context';
 import {
   deleteAllTimelines,
@@ -249,6 +255,56 @@ export default ({ getService }: FtrProviderContext): void => {
             expect(reviewResponse.stats.num_rules_with_conflicts).toBe(1);
             expect(reviewResponse.stats.num_rules_with_non_solvable_conflicts).toBe(0);
           });
+
+          it('should handle long multi-line strings without timing out', async () => {
+            // Install base prebuilt detection rule
+            await createHistoricalPrebuiltRuleAssetSavedObjects(es, [
+              createRuleAssetSavedObject({
+                rule_id: 'rule-1',
+                version: 1,
+                description: TEXT_XL_A,
+              }),
+            ]);
+            await installPrebuiltRules(es, supertest);
+
+            // Customize a multi line string field on the installed rule
+            await patchRule(supertest, log, {
+              rule_id: 'rule-1',
+              description: TEXT_XL_B,
+            });
+
+            // Increment the version of the installed rule, update a multi line string field, and create the new rule assets
+            const updatedRuleAssetSavedObjects = [
+              createRuleAssetSavedObject({
+                rule_id: 'rule-1',
+                version: 2,
+                description: TEXT_XL_C,
+              }),
+            ];
+            await createHistoricalPrebuiltRuleAssetSavedObjects(es, updatedRuleAssetSavedObjects);
+
+            // Call the upgrade review prebuilt rules endpoint and check that one rule is eligible for update
+            // and multi line string field update has no conflict
+            const reviewResponse = await reviewPrebuiltRulesToUpgrade(supertest);
+            expect(reviewResponse.rules[0].diff.fields.description).toEqual({
+              base_version: TEXT_XL_A,
+              current_version: TEXT_XL_B,
+              target_version: TEXT_XL_C,
+              merged_version: TEXT_XL_MERGED,
+              diff_outcome: ThreeWayDiffOutcome.CustomizedValueCanUpdate,
+              merge_outcome: ThreeWayMergeOutcome.Merged,
+              conflict: ThreeWayDiffConflict.SOLVABLE,
+              has_update: true,
+              has_base_version: true,
+            });
+            expect(reviewResponse.rules[0].diff.num_fields_with_updates).toBe(2);
+            expect(reviewResponse.rules[0].diff.num_fields_with_conflicts).toBe(1);
+            expect(reviewResponse.rules[0].diff.num_fields_with_non_solvable_conflicts).toBe(0);
+
+            expect(reviewResponse.stats.num_rules_to_upgrade_total).toBe(1);
+            expect(reviewResponse.stats.num_rules_with_conflicts).toBe(1);
+            expect(reviewResponse.stats.num_rules_with_non_solvable_conflicts).toBe(0);
+          });
         });
 
         describe('when all versions are not mergable', () => {