data-basic: A database library with a focus on ease of use, type safety and useful error messages

[ database, library, mit ] [ Propose Tags ]
Versions [RSS] 0.2.0.0, 0.2.0.1, 0.2.0.2, 0.2.0.3, 0.2.0.4, 0.3.0.0
Change log CHANGELOG.md
Dependencies aeson, attoparsec, base (>=4.7 && <5), binary, bytestring, cases, containers, hssqlppp, lens, lens-aeson, mtl, overload, postgresql-simple, scientific, simple-effects (>=0.10.0.0), simple-logging, string-conv, template-haskell (>=2.12.0.0), text, time [details]
License MIT
Author Nikola Henezi, Luka Horvat
Maintainer nikola@henezi.com, luka.horvat9@gmail.com
Category Database
Home page https://gitlab.com/haskell-hr/basic
Source repo head: git clone https://gitlab.com/haskell-hr/basic.git
Uploaded by darwin226 at 2018-03-08T10:56:53Z
Distributions
Reverse Dependencies 1 direct, 0 indirect [details]
Downloads 4188 total (10 in the last 30 days)
Rating 2.0 (votes: 1) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Docs available [build log]
Last success reported on 2018-03-08 [all 1 reports]

Readme for data-basic-0.3.0.0

[back to package description]

This is a guide on how to get started with basic. Each commit to this repository represents a single step of the guide. We'll cover the motivation and the general description of the library and how to use it, which consists of declaring your schema, generating the model and manipulating the data.

About

Basic is a database library with 4 main objectives, roughly prioritized from first to last.

Ease of use for the most common use cases

While SQL allows for a large number of ways to manipulate your data, people use a small subset of them in the majority of cases.

We aim to make these cases as painless as possible. It should never feel like you need to jump through hoops to get a list of entities that satisfy a simple condition.

Type safety

The type level constraints should reflect your data constraints as best as possible. The library should never allow you to execute a query that doesn't make sense. Expressivity must never come before safety.

Flexibility

While we provide an escape hatch for writing raw queries to ensure you're never "stuck", the cases where it's needed should be few and far in between. If there's a way to provide an elegant and safe solution for a specific query, the library should allow it.

Ability to debug

If you mess up and/or something goes wrong we want to make it as easy as possible to fix it. Legible runtime errors are a must. Also, while Haskell libraries (especially the very "type safe" ones) are notorious for hard to understand compiler errors, we use custom type errors to try and cover some of the standard sources of mistakes. We try to provide useful descriptions of what went wrong and tips on how to fix it.

Getting started

Prerequisites that are not covered by the tutorial:

  • A PostgreSQL installation
  • A database with a desired role configuration
  • A new stack project (simple template) with a dependency on basic that builds

As a database-first library, your model will come from an SQL schema. If you have an existing database, you can use tools like pg_dump to get the schema out. If you're developing a new database, you will declare your tables and relations in an SQL file and then use it to initialize the database and generate the basic model.

Start off by creating a ./model/schema.sql file. We'll be creating a rudimentary blog so we'll need an author and a post table.

CREATE TABLE author (
    id int not null primary key,
    name text not null,
    registration_date timestamp not null
);

CREATE TABLE post (
    id int not null,
    name text not null,
    content text not null,
    creation_date timestamp not null,
    author int not null references author(id)
);

Load the schema into the database you created before. For example like this psql -U postgres -d basic_guide -f model/schema.sql

Now it's time to generate the model. Create a new module in your src directory. Call it Model.

{-# LANGUAGE TemplateHaskell, DataKinds, FlexibleInstances, TypeFamilies, MultiParamTypeClasses #-}
{-# LANGUAGE DeriveGeneric, FlexibleContexts #-}
module Model where

import Data.Basic

mkFromFile "./model/schema.sql"

Add it to your cabal file.

This will generate all the declarations that you'll need to use the library. You will import this module from every module that requires access to the DB. To see the generated declarations, you can open the repl with stack repl and type :browse Model.

Manipulating data

In your main module, import the Model and Data.Basic so we can start playing with data. Let's start by inserting some data. All basic functions work in any monad with a MonadEffect Basic instance. Practically, this means it integrates well with a mtl-style codebase.

sandbox :: MonadEffect Basic m => m ()
sandbox = return ()

For each table in your schema (say, author), basic will generate a new value called newAuthor. This represents a fresh table entry with no data. The underlying value is a record with one field for each column.

Before going futher, add dependencies to the time, mtl, postgresql-simple and lens libraries.

Basic uses the lens approach extensively and setting the fields is no exception. Here's how you do it

{-# LANGUAGE FlexibleContexts, OverloadedStrings #-}
module Main where

import Prelude hiding (id)
import Model
import Data.Basic (MonadEffect, Basic)
import Data.Time (getCurrentTime, getCurrentTimeZone, getCurrentTime, utcToLocalTime)
import Control.Monad.IO.Class (MonadIO(..))
import Data.Function ((&))
import Control.Lens ((.~))

sandbox :: (MonadEffect Basic m, MonadIO m) => m ()
sandbox = do
    zone <- liftIO getCurrentTimeZone
    now <- liftIO getCurrentTime
    newAuthor
        & id .~ 0
        & name .~ "John"
        & registrationDate .~ utcToLocalTime zone now
        & print
        & liftIO

main :: IO ()
main = putStrLn "hello world"

Let's unpack this a bit. We first do a bit of time boilerplate to get the current LocalTime. Then we use the lens syntax to set fields. If you haven't seen it before, the pattern is record & field .~ value. This returns a new record so we can continue chaining. The & operator is actually the flipped version of $ so we use it also to print the final value and lift that IO operation into our monad.

Now, there's no instance MonadEffect Basic IO because the regular old IO monad doesn't know anything about your database. To handle that constraint we use the handleBasicPsql function. It takes a database connection which we can get using the connectPostgreSQL function provided by postgresql-simple. So what we do is handleBasicPsql conn sandbox and it will provide the database functionality that sandbox needs, take care of that constraint leaving only the MonadIO of which the IO monad is an instance. Here's the final version of our Main module.

{-# LANGUAGE FlexibleContexts, OverloadedStrings #-}
module Main where

import Prelude hiding (id)
import Model
import Data.Basic (MonadEffect, Basic, handleBasicPsql)
import Data.Time (getCurrentTime, getCurrentTimeZone, getCurrentTime, utcToLocalTime)
import Control.Monad.IO.Class (MonadIO(..))
import Data.Function ((&))
import Control.Lens ((.~))
import Database.PostgreSQL.Simple

sandbox :: (MonadEffect Basic m, MonadIO m) => m ()
sandbox = do
    zone <- liftIO getCurrentTimeZone
    now <- liftIO getCurrentTime
    newAuthor
        & id .~ 0
        & name .~ "John"
        & registrationDate .~ utcToLocalTime zone now
        & print
        & liftIO

main :: IO ()
main = do
    conn <- connectPostgreSQL "host=localhost port=5432 user=postgres dbname=basic_guide password=admin"
    handleBasicPsql conn sandbox

make sure to modify the connection string

And guess what, we can actually run the thing now! Compile the project and run it, you should see something like:

Entity {_getEntity = Author {_author_id = 0, _author_name = "John", _author_registration_date = 2017-06-09 23:11:16.0531469}}

You might complain that we didn't actually do anything with the database, and you're right, so let's actually insert this author in. How?

    newAuthor
        & id .~ 0
        & name .~ "John"
        & registrationDate .~ utcToLocalTime zone now
        & insert
        & void

add insert to the import list of Data.Basic and import Control.Monad (void)

Doesn't get much simpler than that.

Here's something cool. What if we forgot to set one of the required fields? Turns out, in basic, a partially filled in entity is a legitimate value. You can use the lens syntax to look at it's fields, you can even serialize it to JSON and back. The cool part is that you can't mess up by accessing an undefined field. Basic enforces this at the type level. Go ahead and try removing the & name .~ "John" line. When you try to compile it you should see something like

* Can't insert entity because the required field "name" is not set

Querying data

Let's start by inserting a bunch of rows so we have something to work with.

sandbox :: (MonadEffect Basic m, MonadIO m) => m ()
sandbox = do
    zone <- liftIO getCurrentTimeZone
    now <- liftIO getCurrentTime
    let localNow = utcToLocalTime zone now
    newAuthor
        & id .~ 0
        & name .~ "John"
        & registrationDate .~ localNow
        & insert
        & void

    newAuthor
        & id .~ 1
        & name .~ "Mark"
        & registrationDate .~ localNow
        & insert
        & void

    newAuthor
        & id .~ 2
        & name .~ "Steve"
        & registrationDate .~ localNow
        & insert
        & void

    newPost
        & id .~ 0
        & author .~ 0
        & creationDate .~ localNow
        & content .~ "ABC"
        & name .~ "ABC"
        & insert
        & void

    newPost
        & id .~ 1
        & author .~ 1
        & creationDate .~ localNow
        & content .~ "DEF"
        & name .~ "DEF"
        & insert
        & void

    newPost
        & id .~ 2
        & author .~ 2
        & creationDate .~ localNow
        & content .~ "GHI"
        & name .~ "GHI"
        & insert
        & void

While, to ensure type safety, the types of basic functions are a bit hard to parse, you can follow a simplified mental model. What you do is pretend that every table is just a list of entries. To manipulate the data you use the analog of standard list functions (mapping, filtering, folding...)

Let's get a list of all posts.

posts <- allPosts -- Pretend `allPosts` has the type `[Post]`

How about all posts with an id less than 2?

posts <- dfilter (\p -> (p ^. id) <. (2 :: Int)) allPosts

^. is from lens, dfilter and <. are from Data.Basic

What about sorting them by name? And maybe getting only the first one.

posts <- dtake 1 $ dsortOn (^. name) $ dfilter (\p -> (p ^. id) <. (2 :: Int)) allPosts

Some operations you can't do on lists, like deleting their content. Doing it in basic is easy. Filter the rows you want to delete and call ddelete on the whole thing.

void $ ddelete $ dfilter (\p -> p ^. id ==. (1 :: Int)) allPosts

You can also update the values using the lens syntax like this

void $ dupdate (\p -> p & id .~ (2 :: Int)) $ dfilter (\p -> p ^. id ==. (1 :: Int)) allPosts

Want to find the highest id of all posts?

liftIO . print =<< dfoldMap (\p -> Max (p ^. id)) allPosts

Things get more interesting with joins and groups. To join two tables you use the djoin function at the start and then pretend your working with a list of all possible pairs. For grouping, you use the dgroupOn function to choose which field(s) to group on. Pretend the signature is something like dgroupOn :: (a -> b) -> [a] -> [(b, [a])]. You can then either map over that and fold the inner list (dmap and dfoldMap) or use the dfoldMapInner convenience functions with the pretend type signature dfoldMapInner :: Monoid m => (a -> m) -> [(b, [a])] -> [(b, m)].

res <-
    dfoldMapInner (\(p, a) -> Max (p ^. id)) $
    dgroupOn (\(p, a) -> a ^. id) $
    dfilter (\(p, a) -> p ^. author ==. a ^. id) $
    djoin allPosts allAuthors

Join all posts and authors, then filter the list so you're left with pairs (post, author of the post). Then group them by the author and, for each author, get it's highest post id.

Entities also know if they're fresh or they came from the database. This let's you conveniently update single rows like this

[a] <- dfilter (\a -> a ^. id ==. (0 :: Int)) allAuthors
a & name .~ "New name"
  & save
  & void

What it does is it looks for the database entry with the same primary key and updates all the other fields. This works if the table actually has a primary key.