Quickly search for code snippets from wherever you are, no need to switch to the browser and disrupt your flow:

As a prerequisite, you will need to have Python 3 installed, then you have some options

You can use Codesearch inside of VS Code, simply install the extension

This is the best way to use Codesearch, it allows you to search with a shortcut from anywhere on the computer, but you will need to have Alfred with Powerpack (£29 one time payment)

It also requires a bit more setup, first, clone this repo and install the dependencies:

git clone https://github.com/rogeriochaves/codesearch.git
cd codesearch
pip3 install -r requirements.txt

Then you have three methods:

This approach is the quickest way to get started, it has the advantage of everything local, no registration needed anywhere and no warmup needed after starting the server. The disadvantage is that you have to download a ~500MB deep learning model, plus leave a server running on the background which takes ~1GB of RAM

Try it out immediately (it will download the model on the first run):

python3 codesearch/cli.py bash replace text in file

The CLI initializes the model each time you run it, to leave it running for faster searches, start the server:

python3 codesearch/server.py

Now in another terminal you can curl it anytime:

curl "localhost:2633/?q=bash+replace+text+in+file"

Finally, import the Codesearch.alfredworkflow to your Alfred and you are done!

To use it, press Option + Space and start searching, or type "how " on regular alfred search box

To always have the server running in the background after restarting the computer, you can go to Mac Preferences > Users & Groups > Login Items tab, click + and add the codesearch/CodesearchServer.app file

This approach has the advantage of not needing to leave a server running and not eating up your RAM, but it will require ~10s warmup time from Hugginface after a period of no use

To set it up, first create an account at Huggin Face and get your API Token.

To confirm it's working, set the API_TOKEN env var and try the CLI:

export API_TOKEN=huggingface_token_here
python3 codesearch/cli.py bash replace text in file

Now, import Codesearch.alfredworkflow to your Alfred, then go to Alfred Preferences > Workflows > Codesearch and click on the [X] button on the top-right, then on the Environment Variables section add two env variables:

API_TOKEN: paste your hugging faces API Token
CODESEARCH_PATH: /path/where/you/cloned/codesearch

If you have a server available you can leave codesearch running there, this has the advantage of not needing warmup time from Hugginface, but the cost of leave a server running. There is a Dockerfile at the repo to facilitate the integration.

Then, for Alfred to work with it, just like on the Method B, add this env variable to the workflow:

SERVER_URL: https://yourserver.com/

Have fun!

When you search a query, in the background we do a Google search and scrape the page for the search results texts. The extracted text is then passed to a Named Entity Recognition Deep Learning model created by @mrm8488 trained on a Stackoverflow corpus by Tabassum et al, and the results are displayed to you.

In summary, we try to simulate what developers do a lot: quickly google something and skim over the results looking for code, then alt+tab back to write it down. Sometimes it's not enough and you need to dig deeper into the search results (that's why Codesearch also suggest to open the browser for a full search with DuckDuckGo) but more often than not a quick look is enough to trigger your memory and get you back on the flow.