git clone ''

(ql:quickload :krisajenkins.yesql)

Yesql - Clojure & SQL rethought.

Yesql is a Clojure library for using SQL.

Build Status


Frozen. Maintainer sought.

Tested with Clojure 1.5-1.9-alpha20, but there will be no new development unless a maintainer steps up.

(I've been promising myself for ages that I'll get round to all the feature requests the next time I'm working on a Clojure/SQL project. But that's been a long while now, so maybe it's time to admit that this project needs a more active pair of hands. If you'd like to take it on, please contact me.

(You might also consider hugsql which is philosophically similar and actively maintained.)


Add this to your Leiningen :dependencies:

Clojars Project


Plus you'll want a database driver. Here are some examples (but double check, because there may be a newer version available):

|Database|:dependencies Entry| |—|—| |PostgreSQL|[org.postgresql/postgresql "9.4-1201-jdbc41"]| |MySQL|[mysql/mysql-connector-java "5.1.32"]| |Oracle|[ ""]| |SQLite|[org.xerial/sqlite-jdbc "3.7.2"]| |Derby|[org.apache.derby/derby ""]| |h2|[com.h2database/h2 "1.4.191"]|

(Any database with a JDBC driver should work. If you know of a driver that's not listed here, please open a pull request to update this section.)

Migrating From Previous Versions

See the Migration Guide.


You're writing Clojure. You need to write some SQL.

I think we're all agreed that this is a problem:

(query "SELECT * FROM users WHERE country_code = ?" "GB")

Unless these query strings are short, they quickly get hard to read and hard to rewrite. Plus the lack of indentation & syntax highlighting is horrible.

But something like this is not the solution:

(select :*
        (from :users)
        (where (= :country_code "GB")))

Clojure is a great language for writing DSLs, but we don't need a new one. SQL is already a mature DSL. And s-expressions are great, but here they're not adding anything. This is parens-for-parens sake. (Don't agree? Wait until this extra syntax layer breaks down and you start wrestling with a (raw-sql) function.)

So what's the solution? Keep the SQL as SQL. Have one file with your query:

-- name: users-by-country
FROM users
WHERE country_code = :country_code

…and then read that file to turn it into a regular Clojure function:

(defqueries "some/where/users_by_country.sql"
   {:connection db-spec})

;;; A function with the name `users-by-country` has been created.
;;; Let's use it:
(users-by-country {:country_code "GB"})
;=> ({:name "Kris" :country_code "GB" ...} ...)

By keeping the SQL and Clojure separate you get:

When Should I Not Use Yesql?

When you need your SQL to work with many different kinds of database at once. If you want one complex query to be transparently translated into different dialects for MySQL, Oracle, Postgres etc., then you genuinely do need an abstraction layer on top of SQL.


One File, One Query

Create an SQL query. Note we can supply named parameters (in snake_case) and a comment string:

-- Counts the users in a given country.
SELECT count(*) AS count
FROM user
WHERE country_code = :country_code

Make sure it's on the classpath. For this example, it's in src/some/where/. Now we can use it in our Clojure program.

(require '[yesql.core :refer [defquery]])

; Define a database connection spec. (This is standard
(def db-spec {:classname "org.postgresql.Driver"
              :subprotocol "postgresql"
              :subname "//localhost:5432/demo"
              :user "me"})

; Import the SQL query as a function.
(defquery users-by-country "some/where/users_by_country.sql"
   {:connection db-spec})

Lo! The function has been created, with automatic, useful docstrings in the REPL:

(clojure.repl/doc users-by-country)

;=> -------------------------
;=> user/users-by-country
;=> ([{:keys [country_code]}]
;=>  [{:keys [country_code]} {:keys [connection]}])
;=>   Counts the users in a given country.

Now we can use it:

; Use it standalone.
(users-by-country {:country_code "GB"})
;=> ({:count 58})

; Use it in a transaction.
(require '[ :as jdbc])

(jdbc/with-db-transaction [tx db-spec]
   {:limeys (users-by-country {:country_code "GB"} {:connection tx})
    :yanks  (users-by-country {:country_code "US"} {:connection tx})})

One File, Many Queries

As an alternative to the above, you can have many SQL statements in a single SQL file. The file format is: (<name tag> [docstring comments] <the query>)*, like so:

-- name: users-by-country
-- Counts the users in a given country.
SELECT count(*) AS count
FROM user
WHERE country_code = :country_code

-- name: user-count
-- Counts all the users.
SELECT count(*) AS count
FROM user

Then read the file in like so:

(require '[yesql.core :refer [defqueries]])
(defqueries "some/where/queryfile.sql"
   {:connection db-spec})

defqueries returns a sequence of the vars it binds, which can be useful feedback while developing.

As with defquery, each function will have a docstring based on the comments, and a parameter map based on the SQL parameters.

? Parameters

Yesql supports named parameters, and ?-style positional parameters. Here's an example:

-- name: young-users-by-country
FROM user
  country_code = ?
  country_code = ?
AND age < :max_age

Supply the ? parameters as a vector under the :? key, like so:

(young-users-by-country {:? ["GB" "US"]
                         :max_age 18})

Selectively import queries

Similarly to defqueries, require-sql lets you create a number of query functions at a time, but with a syntax more like clojure.core/require.

Using the queryfile.sql from the previous example:

(require '[yesql.core :refer [require-sql]])

; Use :as to alias the entire namespace, and :refer to refer functions
; into the current namespace. Use one or both.
(require-sql ["some/where/queryfile.sql" :as user :refer [user-count])

;=> ({:count 132})

(user/users-by-country db-spec "GB")
;=> ({:count 58})

IN-list Queries

Yesql supports IN-style queries. Define your query with a single-element in the IN list, like so:

-- name: find-users
-- Find the users with the given ID(s).
FROM user
WHERE user_id IN (:id)
AND age > :min_age

And then supply the IN-list as a vector, like so:

(defqueries "some/where/queryfile.sql"
   {:connection db-spec})

(find-users {:id [1001 1003 1005]
             :min_age 18})

The query will be automatically expanded to ... IN (1001, 1003, 1005) ... under the hood, and work as expected.

Just remember that some databases have a limit on the number of values in an IN-list, and Yesql makes no effort to circumvent such limits.

Row And Result Processors

Like, Yesql accepts functions to pre-process each row, and the final result, like so:

-- name: current-time
-- Selects the current time, according to the database.
SELECT sysdate
FROM dual;
(defqueries "/some/where/queryfile.sql"
  {:connection db-spec})

;;; Without processors, this query returns a list with one element,
;;;   containing a map with one key:
;=> ({:sysdate #inst "2014-09-30T07:30:06.764000000-00:00"})

;;; With processors we just get the value we want:
(current-time {} {:result-set-fn first
                  :row-fn :sysdate
                  :identifiers identity})
;=> #inst "2014-09-30T07:30:06.764000000-00:00"

As with the default :result-set-fn is doall, the default :row-fn is identity, and the default :identifiers is clojure.string/lower-case.

A note of caution: Remember you're often better off doing your processing directly in SQL. For example, if you're counting a million rows, you can do it with {:result-set-fn count} or SELECT count(*) .... Both wil give the same answer, but the SQL-version will avoid sending a million rows over the wire to do it.

Insert/Update/Delete and More

To do INSERT/UPDATE/DELETE statements, you just need to add an ! to the end of the function name, and Yesql will execute the function appropriately. For example:

-- name: save-person!
UPDATE person
SET name = :name
WHERE id = :id
(save-person! {:id 1
               :name "Dave"})
;=> 1

A !-tagged function will return the number of rows affected.

! enables every statement type - not just INSERT/UPDATE/DELETE but also CREATE/DROP/ALTER/BEGIN/... - anything your driver will support.

Insert, Returning Autogenerated Keys

There's one more variant: when you want to insert data and get back a database-generated primary key, the driver requires a special call, so Yesql needs to be specially-informed. You can do an “insert returning autogenerated key” with the <! suffix, like so:

-- name: create-person<!
INSERT INTO person (name) VALUES (:name)
(create-person<! {:name "Dave"})
;=> {:name "Dave" :id 5}

The exact return value will depend on your database driver. For example PostgreSQL returns the whole row, whereas Derby returns just {:1 5M}.

The <! suffix is intended to mirror core.async, so it should be easy to remember.

Development & Testing

Yesql uses the marvellous expectations library for tests. It's like clojure.test, but has lighter-weight syntax and much better failure messages.

Call lein test to run the test suite. Call lein test-all to run the tests against all (supported) versions of Clojure. Call lein autoexpect to automatically re-run the tests as source files change.

Other Languages

Yesql has inspired ports to other languages:

|Language|Project| |—|—| |JavaScript|Preql| |JavaScript|sqlt| |Python|Anosql| |Go|DotSql| |Go|goyesql| |C#|JaSql| |Ruby|yayql| |Erlang|eql| |Clojure|YeSPARQL| |PHP|YepSQL|


Copyright © 2013-2016 Kris Jenkins

Distributed under the Eclipse Public License, the same as Clojure.

PS - Is Yesql An ORM?

No. There are no Objects here, only Values. Yesql is a VRM. This is better because it's pronounced, “Vroom!”