Node specific code been bundled only in production mode (tesseract.js)

[yue4u]

⚠️ IMPORTANT ⚠️ Please do not ignore this template. If you do, your issue will be closed immediately.

• Read the docs.
• Make sure this is a Vite issue and not a framework-specific issue. For example, if it's a Vue SFC related bug, it should likely be reported to https://github.com/vuejs/vue-next instead.
• This is a concrete bug. For Q&A open a GitHub Discussion or join our Discord Chat Server.

Describe the bug

It works fine when using tesseract.js as a dependency in development mode, but throws error in production build. I did some investigation and found it looks like the browser field in package.json does not map correctly in production build.

browser field: https://github.com/naptha/tesseract.js/blob/90466c3b5504a9220ba0ff91ccec22003f72cbd2/package.json#L29-L31
where it was loaded: https://github.com/naptha/tesseract.js/blob/90466c3b5504a9220ba0ff91ccec22003f72cbd2/src/createWorker.js#L14

This can be temporally fixed via patch-package to the tesseract.js package on the line where it switches import. Wondering if this could be fixed on the vite side or by adding extra configuration.

Reproduction

https://github.com/rainy-me/vite-tesseract-reproduction

System Info

• vite version: v2.0.5
• Operating System: Linux
• Node version: v15.8.0
• Package manager (npm/yarn/pnpm) and version: yarn 1.22.10

Logs (Optional if provided reproduction)

1. Run vite or vite build with the --debug flag.
2. Provide the error log here.

$ vite build --debug
  vite:config no config file found. +0ms
  vite:config using resolved config: {
  vite:config   root: '/home/yue/dev/vite-tesseract-reproduction',
  vite:config   base: '/',
  vite:config   mode: 'production',
  vite:config   configFile: undefined,
  vite:config   logLevel: undefined,
  vite:config   clearScreen: undefined,
  vite:config   build: {
  vite:config     target: [ 'es2019', 'edge16', 'firefox60', 'chrome61', 'safari11' ],
  vite:config     polyfillDynamicImport: true,
  vite:config     outDir: 'dist',
  vite:config     assetsDir: 'assets',
  vite:config     assetsInlineLimit: 4096,
  vite:config     cssCodeSplit: true,
  vite:config     sourcemap: false,
  vite:config     rollupOptions: {},
  vite:config     commonjsOptions: { include: [Array], extensions: [Array] },
  vite:config     minify: 'terser',
  vite:config     terserOptions: {},
  vite:config     cleanCssOptions: {},
  vite:config     write: true,
  vite:config     emptyOutDir: null,
  vite:config     manifest: false,
  vite:config     lib: false,
  vite:config     ssr: false,
  vite:config     ssrManifest: false,
  vite:config     brotliSize: true,
  vite:config     chunkSizeWarningLimit: 500
  vite:config   },
  vite:config   inlineConfig: {
  vite:config     root: undefined,
  vite:config     base: undefined,
  vite:config     mode: undefined,
  vite:config     configFile: undefined,
  vite:config     logLevel: undefined,
  vite:config     clearScreen: undefined,
  vite:config     build: {}
  vite:config   },
  vite:config   resolve: { dedupe: undefined, alias: [ [Object] ] },
  vite:config   publicDir: '/home/yue/dev/vite-tesseract-reproduction/public',
  vite:config   command: 'build',
  vite:config   isProduction: true,
  vite:config   optimizeCacheDir: '/home/yue/dev/vite-tesseract-reproduction/node_modules/.vite',
  vite:config   plugins: [
  vite:config     'alias',
  vite:config     'vite:dynamic-import-polyfill',
  vite:config     'vite:resolve',
  vite:config     'vite:html',
  vite:config     'vite:css',
  vite:config     'vite:esbuild',
  vite:config     'vite:json',
  vite:config     'vite:wasm',
  vite:config     'vite:worker',
  vite:config     'vite:asset',
  vite:config     'vite:define',
  vite:config     'vite:css-post',
  vite:config     'vite:build-html',
  vite:config     'commonjs',
  vite:config     'vite:data-uri',
  vite:config     'rollup-plugin-dynamic-import-variables',
  vite:config     'vite:import-analysis',
  vite:config     'vite:esbuild-transpile',
  vite:config     'vite:terser',
  vite:config     'vite:reporter'
  vite:config   ],
  vite:config   server: {},
  vite:config   env: { BASE_URL: '/', MODE: 'production', DEV: false, PROD: true },
  vite:config   assetsInclude: [Function: assetsInclude],
  vite:config   logger: {
  vite:config     hasWarned: false,
  vite:config     info: [Function: info],
  vite:config     warn: [Function: warn],
  vite:config     error: [Function: error],
  vite:config     clearScreen: [Function: clearScreen]
  vite:config   },
  vite:config   createResolver: [Function: createResolver]
  vite:config } +2ms
vite v2.0.5 building for production...
✓ 72 modules transformed.
dist/assets/favicon.17e50649.svg   1.49kb
dist/index.html                    0.52kb
dist/assets/index.ccce2ca3.css     0.16kb / brotli: 0.10kb
dist/assets/index.07b4c97f.js      2.21kb / brotli: 1.51kb
dist/assets/vendor.6502411a.js     81.76kb / brotli: 24.03kb

[yongjun21]

Looks like there are a couple of issues related to how Vite handles the browser field in package.json.

#1979
#2320

[yue4u]

After several updates(vite and other libs) the fix above doesn't work anymore and looks like vite is not bunlding dynamic imported worker or wasm?

Anyway I found another hacky way to fix this issue: directly importing them via ?url will do the trick, no need for patch-package anymore.

import { createWorker } from "tesseract.js";
import workerPath from "tesseract.js/dist/worker.min.js?url";
import corePath from "tesseract.js-core/tesseract-core.wasm.js?url";

createWorker({
  workerPath,
  corePath,
  // ..
})

I'm closing this issue since I'm not sure if the origin problem still exists.