fix(stega): 2x faster encoding of portable text (#920)

.github/workflows/ci.yml

@@ -30,6 +30,7 @@ jobs:
         run: echo "::add-matcher::.github/eslint-compact.json"
       - run: npm run lint -- --report-unused-disable-directives -f compact
       - run: npm run coverage
+      - run: npm run benchmark
       - uses: actions/upload-artifact@v4
         name: Cache build output
         with:

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sanity/client",
-  "version": "6.22.4",
+  "version": "6.22.5-canary.0",
   "description": "Client for retrieving, creating and patching data from Sanity.io",
   "keywords": [
     "sanity",
@@ -89,6 +89,7 @@
     "umd"
   ],
   "scripts": {
+    "benchmark": "npm test -- bench",
     "prebuild": "npm run clean",
     "build": "pkg build --strict && pkg --strict && npm run rollup && npm run minify",
     "clean": "npx rimraf dist coverage umd/*.js",
@@ -126,6 +127,7 @@
     "@edge-runtime/vm": "^4.0.4",
     "@rollup/plugin-commonjs": "^28.0.1",
     "@rollup/plugin-node-resolve": "^15.3.0",
+    "@sanity/client-latest": "npm:@sanity/client@latest",
     "@sanity/pkg-utils": "^6.11.11",
     "@types/json-diff": "^1.0.3",
     "@types/node": "^22.9.0",

src/csm/walkMap.ts

@@ -25,6 +25,17 @@ export function walkMap(
   }
 
   if (isRecord(value)) {
+    // Handle Portable Text in a faster way
+    if (value._type === 'block' || value._type === 'span') {
+      const result = {...value}
+      if (value._type === 'block') {
+        result.children = walkMap(value.children, mappingFn, path.concat('children'))
+      } else if (value._type === 'span') {
+        result.text = walkMap(value.text, mappingFn, path.concat('text'))
+      }
+      return result
+    }
+
     return Object.fromEntries(
       Object.entries(value).map(([k, v]) => [k, walkMap(v, mappingFn, path.concat(k))]),
     )

src/stega/filterDefault.ts

@@ -17,30 +17,6 @@ export const filterDefault: FilterDefault = ({sourcePath, resultPath, value}) =>
     return false
   }
 
-  /**
-   * Best effort infer Portable Text paths that should not be encoded.
-   * Nothing is for certain, and the below implementation may cause paths that aren't Portable Text and otherwise be safe to encode to be skipped.
-   * However, that's ok as userland can always opt-in with the `encodeSourceMapAtPath` option and mark known safe paths as such, which will override this heuristic.
-   */
-  // If the path ends in marks[number] it's likely a PortableTextSpan: https://github.com/portabletext/types/blob/e54eb24f136d8efd51a46c6a190e7c46e79b5380/src/portableText.ts#LL154C16-L154C16
-  if (typeof endPath === 'number' && sourcePath.at(-2) === 'marks') {
-    return false
-  }
-  // Or if it's [number].markDefs[number].href it's likely a PortableTextLink: https://github.com/portabletext/types/blob/e54eb24f136d8efd51a46c6a190e7c46e79b5380/src/portableText.ts#L163
-  if (
-    endPath === 'href' &&
-    typeof sourcePath.at(-2) === 'number' &&
-    sourcePath.at(-3) === 'markDefs'
-  ) {
-    return false
-  }
-  // Otherwise we have to deal with special properties of PortableTextBlock, and we can't confidently know if it's actually a `_type: 'block'` array item or not.
-  // All we know is that if it is indeed a block, and we encode the strings on these keys it'll for sure break the PortableText rendering and thus we skip encoding.
-  // https://github.com/portabletext/types/blob/e54eb24f136d8efd51a46c6a190e7c46e79b5380/src/portableText.ts#L48-L58
-  if (endPath === 'style' || endPath === 'listItem') {
-    return false
-  }
-
   // Don't encode into anything that is suggested it'll render for SEO in meta tags
   if (
     sourcePath.some(

test/stega/__snapshots__/stegaEncodeSourceMap.test.ts.snap

@@ -73,17 +73,14 @@ exports[`resolveEditUrl '/' > logger.log 1`] = `
     "[@sanity/client]: Encoding source map into result",
   ],
   [
-    "[@sanity/client]: Paths encoded: 18, skipped: 81",
+    "[@sanity/client]: Paths encoded: 18, skipped: 36",
   ],
   [
     "[@sanity/client]: Table of encoded paths",
   ],
   [
     "[@sanity/client]: List of skipped paths",
     [
-      "description[]._key",
-      "description[]._type",
-      "description[].style",
       "slug.current",
       "slug._type",
       "media[]._type",
@@ -227,17 +224,14 @@ exports[`resolveEditUrl 'https://test.sanity.studio' > logger.log 1`] = `
     "[@sanity/client]: Encoding source map into result",
   ],
   [
-    "[@sanity/client]: Paths encoded: 18, skipped: 81",
+    "[@sanity/client]: Paths encoded: 18, skipped: 36",
   ],
   [
     "[@sanity/client]: Table of encoded paths",
   ],
   [
     "[@sanity/client]: List of skipped paths",
     [
-      "description[]._key",
-      "description[]._type",
-      "description[].style",
       "slug.current",
       "slug._type",
       "media[]._type",

test/stega/stegaEncodeSourceMap.bench.ts

@@ -0,0 +1,20 @@
+import {stegaEncodeSourceMap as stegaEncodeSourceMapLatest} from '@sanity/client-latest/stega'
+import {bench, describe} from 'vitest'
+
+import {stegaEncodeSourceMap} from '../../src/stega/stegaEncodeSourceMap'
+import type {InitializedStegaConfig} from '../../src/stega/types'
+import data from './stegaSnapshotHuge.json' with {type: 'json'}
+
+const config = {
+  enabled: true,
+  studioUrl: 'https://test.sanity.studio',
+} satisfies InitializedStegaConfig
+
+describe('stegaEncodeSourceMap can handle Portable Text in a performant way', () => {
+  bench('@sanity/client@latest', () => {
+    stegaEncodeSourceMapLatest(data.result, data.resultSourceMap as any, config)
+  })
+  bench('src', () => {
+    stegaEncodeSourceMap(data.result, data.resultSourceMap as any, config)
+  })
+})