# Fragment store Quilt stores local package state in a directory `quilt_packages` in your system's user data directory. To find the location of your `quilt_packages` directory, run: ```bash quilt ls ``` ## Directory Structure ```bash BASE_DIR/quilt_packages ``` Quilt stores local copies of packages in a single directory, `quilt_packages`. That directory contains both the raw data and metadata for packages including any tags and versions associated with each package instance. Using files in the file system to store quilt metadata is perhaps slower than in a database or single-file lookup, but individual files are more human-readable, require fewer dependencies to manipulate, and are easier to repair if they get corrupted. Concurrent access to quilt metadata is also easier to manage. ### Fragments ```bash quilt_packages/objs/ ``` Stores binary data fragments (a.k.a. "objects") identified by hash. These fragments include compressed raw files and Parquet files. Fragment hashes are verified and fragments are stored only once (de-duplication). ### Contents ```bash quilt_packages///contents/ ``` The per-package contents directory `contents` uses the file system to store a mapping of package names, versions and tags to locally installed package instances. ### Tags and Versions ```bash quilt_packages///tags/ : quilt_packages///versions/ : ``` ## Implementation Sketch High-level outlines of how basic Quilt commands interact with local storage. ### build * parse build.yml and iterate through package tree: * create binary fragments and save to objs directory * build in-memory package tree * calculate package instance hash * save package manifest (identified by hash) * add repo:latest tag ### push * find instance: * by pkg/tag * lookup hash in pkgname/tags/tag->hash in contents * by pkg/version * lookup hash in pkgname/versions/version->hash in contents * by hash (direct lookup) * upload binary fragments * post compressed package manifest to registry ### install * lookup pkg in contents to check for local copy * warn if present and not force * lookup hash from registry * download package manifest * verify package hash against contents * save by package manifest (by package hash) * download and save binary fragments (if not present) ### ls * read all local package instances and build map of metadata: * hash: (size, created, etc.) * read contents and sort by package * foreach package: * lookup versions: * list all instances with versions, ordered by version * lookup tags: * list all instances with tags, ordered by tag * alternate: * order instances by reverse creation date * mark each instance printed in metadata map * list all un-tagged, un-versioned instances for the package ### import * lookup package hash in contents * by tag: package/tags/tag->hash * by version: package/versions/version->hash * by hash (verify hash is associated with package) * read package manifest (find by hash) * parse package tree --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.quilt.bio/quilt-2-master/developer/repo-format.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.