A simple script to be used in Dockerfile to do yarn install without copying all package.json
s in a monorepo workspace. Use case inspired by pnpm's fetch command but for Yarn Berry.
# ------------------------------------------------------------------------------
# First image (install dependencies)
# ------------------------------------------------------------------------------
FROM node:16-alpine as yarn
LABEL stage=yarn
WORKDIR /workspace
COPY .yarn ./.yarn
COPY yarn.lock .yarnrc.yml ./
RUN yarn dlx yarn-lock-to-package-json
RUN yarn install --immutable
# ------------------------------------------------------------------------------
# Server image (build image with production dependencies only)
# ------------------------------------------------------------------------------
FROM yarn as server
LABEL stage=server
WORKDIR /workspace
# Overwrite dummy package.json with actual ones (so that we can use scripts)
COPY package.json ./
COPY packages ./packages
COPY apps/server ./apps/server
RUN yarn workspaces foreach -tR --from @workspace-name/server run build
RUN cd apps/server && yarn workspaces focus --production
CMD ["yarn", "workspace", "@workspace-name/server", "start"]
For similar functionality but packaged in a yarn plugin you can check out yarn-plugin-fetch by @devthejo, it uses this package internally to provide the same feature plus adds workspace focus support and more tools.
What it does is parse the lockfile and recreate the monorepo structure by creating dummy package.json
for each package in the workspace. These dummy package.json don't invalidate your docker cache when you change a script or increment a version and so your docker build cache remains valid up to the layer of RUN yarn install --immutable
.
Test it by copying your yarn.lock
to a new folder and run npx yarn-lock-to-package-json
.
This is the dummy file generated:
{
"name": "@monorepo/package",
"version": "0.0.0",
"description": "**DON'T COMMIT** Generated file for caching",
"private": true,
"dependencies": {
"typescript": "^4.5.5"
}
}
These existing issues in yarn