Skip to content

A simple pastebin/gist like service (server side), with git as its backend

License

Notifications You must be signed in to change notification settings

amirgon/snippetbin

Repository files navigation

SnippetBin


Rational

A simple pastebin/gist-like service (server side). The underlying database is git. Syncs automatically with GitHub or any other git server.

It allows

  • Creating new (unamed) files and saving them on the server
  • A file revision is immutable. Once written to the server it cannot be changed.
  • A file revision can have history. It is an edit of another revision.
  • Sharing the files by publishing a "revision id" that identifies that revision of that file
  • Editing an existing file revision according to its revision id (this will create a new file with the original text and original history)
  • Browsing the change history of a file, and branching it to a new file

Configuration and deployment

To deploy in a Docker container, first build the docker image:

sudo docker build -t snippetbin ./

Then run it with environemnt variables like this:

sudo docker run -d --restart unless-stopped --name snippetbin -p 443:443 \
        -v /var/snippetbin/data:/snippetbin_data \
        -e "SNIPPETBIN_PORT=443" \
        -e "SNIPPETBIN_DEPLOY_KEY=$(cat /var/snippetbin/deploy_key)" \
        -e "SNIPPETBIN_REMOTE_REPO=my_snippetbin_data.git" \
        -e "SNIPPETBIN_DATA_DIR=/snippetbin_data" \
        -e "SNIPPETBIN_SSL_KEY=$(cat /var/snippetbin/ssl.key)" \
        -e "SNIPPETBIN_SSL_CERT=$(cat /var/snippetbin/ssl.cer)" \
        snippetbin npm start

This example would start SnippetBin in Docker and start HTTPS server on port 443, with the specified GitHub deploy key (ssh) and certificates (ssl).
The data directory would be mounted to /var/snippetbin/data on the host.
It will look for the deploy_key and ssl keys under /var/snippetbin/ on the host.
It will use my_snippetbin_data.git repo on as backing database.

Demo

https://snippet-bin.herokuapp.com

API

LOAD-FILE

Input:

  • revision key

Output:

  • file text
  • revision history (list of revision keys)

SAVE-FILE

Input:

  • file text
  • original revision key (optional)

Output:

  • new revision key

Client flows

New file

  • User provides new text, and asks to save the file
  • Client calls SAVE-FILE with user text.
  • Client publishes revision key in a URL
  • Client saves revision key. It will be used when the user edits the file

Edit existing file

  • User continues to edit an existing file, and asks to save the file. (or file automatially saved)
  • Client remembers the revision key for the file the user is editing
  • Client calls SAVE-FILE with user text and the revision key

Load file

  • User provides revision key (through URL, or by selecting it from revision history)
  • Client calls LOAD-FILE with the revision key
  • Client displays file text to the user, and allows editing it.
  • Client displays the user the revision history, in case the user wants to go back to historic revision of the file
  • Client saves revision key. It will be used when the user saves his edits of the file

Implementation

  • Underlying database is git.
  • Revision key is a commit hash
  • Revision history is the log of a given commit hash
  • Each commit will contain a singe file
  • When saving a file
    • A new file without "original revision key" will be named by a unique ID and commited.
    • An existing file with commit ID will be branched from the original revision ID.

NOTE: each commit must have a parent, otherwise git diff-tree will not work correctly. So don't start with empty repo!

git commands for LOAD-FILE

Get file text according to revision key. Revision key is actually the commit ID, and each commit has only one file

Inputs:

  • revision_key

Outputs:

  • file_text
  • revision_history
file_hash=$(git diff-tree --no-commit-id "${revision_key}" | cut -d ' ' -f 4)
file_text=$(git cat-file -p "${file_hash}")
file_name=$(git diff-tree --no-commit-id --name-only "${revision_key}")
revision_history=($(git rev-list "${revision_key}" -- "${file_name}"))

git commands for SAVE-FILE

Inputs:

  • file_text
  • optional original_revision_key
  • commit_msg

Outputs:

  • revision_key
if [ -z "$original_revision_key" ]
then
    file_name=$(uuidgen)
    git checkout master
else
    file_name=$(git diff-tree --no-commit-id --name-only -r "$original_revision_key")
    git checkout -B branch_$(uuidgen) "$original_revision_key"
fi
echo "$file_text" > "$file_name"
git add "$file_name"
git commit -am "$commit_msg"
revision_key=$(git rev-parse HEAD)

TODO

  • Change SAVE-FILE implementation to use git plumbing commands
  • Prevent checking out, no need to update all files locally when only one is used simultaneously

About

A simple pastebin/gist like service (server side), with git as its backend

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published