Stackage is a neat tool to use when developing Software in Haskell. It’s integrated with the Hoogle search engine that can search on type signatures (and function names) which lets you find almost any function easily.
One of the downsides of Hoogle is that it can be slow for people with some slower connections and so looking up any kind of documentation becomes a really frustrating experience.
The other issue with using Hoogle is that when you look up documentation for a function of a library that you use in a project, you need to make sure you are looking at the documentation for the exact version of that library that you are using in your package.yml or .cabal file; otherwise you maybe looking at an entirely different api and end up wasting a lot of time.
Stackage solves this by letting you search against a particular snapshot where all the versions of your libraries will be pinned against that snapshot (see the resolver field in stackage.yaml).
Wouldn’t it be nice if we could run Hoogle locally for our project with all the documentation we need available at our fingertips? What would be even better is that Hoogle was seeded with only the versions of dependencies in our project that we care about.
Well thanks to the fantastic “An opinionated guide to Haskell in 2018” by Alexis King we now know how!
The basic steps are:
Generate Haddock documentation when you compile your project
stack test --fast --haddock-deps
The first time you run this it can take a while.
Generate a local Hoogle database
stack hoogle -- generate --local
Run your local Hoogle server
stack hoogle -- server --local --port=8080
Profit
You can get full access to your local Hoogle by opening localhost:8080
in your browser.
Something to be aware of is that when you add new dependencies or change existing ones in your project you need to generate your local Hoogle database again:
Unfortunately, you will have to manually regenerate the Hoogle database
when you install new packages and their documentation, which you can do
by re-running stack hoogle -- generate --local.
Fortunately, regenerating the database doesn’t take very long, as long
as you’ve been properly rebuilding the documentation with
--haddock-deps.
This shouldn’t take too long though if you’ve been generating your Haddock as per step 1.
Original post can be found here.