git clone ''

(ql:quickload :caveman)

Caveman2 - Lightweight web application framework

Build Status


(defparameter *web* (make-instance '<app>))

@route GET "/"
(defun index ()
  (with-layout (:title "Welcome to My site")
    (render #P"index.tmpl")))

@route GET "/hello"
(defun say-hello (&key (|name| "Guest"))
  (format nil "Hello, ~A" |name|))

About Caveman2

What's the difference from Caveman “1”?

All of them. Caveman2 was written from scratch.

These are noticeable points.

The reason I wrote it from scratch

One of the most frequently asked questions was “Which should I use ningle or Caveman? What are the differences?” I think it was because the roles of them were too similar. Both of them are saying “micro” and no database support.

Caveman2 is no more “micro” web application framework. It supports CL-DBI and has database connection management by default. Caveman has started growing up.

Design Goal

Caveman is intended to be a collection of common parts of web applications. Caveman has 3 rules to make decisions.


You came here because you're interested in living like a caveman, right? There's no Disneyland, but it's good place to start. Let's get into a cave.


Caveman2 is now available on Quicklisp.

(ql:quickload :caveman2)

Generating a project skeleton

(caveman2:make-project #P"/path/to/myapp/"
                       :author "<Your full name>")
;-> writing /path/to/myapp/.gitignore
;   writing /path/to/myapp/README.markdown
;   writing /path/to/myapp/app.lisp
;   writing /path/to/myapp/db/schema.sql
;   writing /path/to/myapp/shlyfile.lisp
;   writing /path/to/myapp/myapp-test.asd
;   writing /path/to/myapp/myapp.asd
;   writing /path/to/myapp/src/config.lisp
;   writing /path/to/myapp/src/db.lisp
;   writing /path/to/myapp/src/main.lisp
;   writing /path/to/myapp/src/view.lisp
;   writing /path/to/myapp/src/web.lisp
;   writing /path/to/myapp/static/css/main.css
;   writing /path/to/myapp/t/myapp.lisp
;   writing /path/to/myapp/templates/_errors/404.html
;   writing /path/to/myapp/templates/index.tmpl
;   writing /path/to/myapp/templates/layout/default.tmpl


Caveman2 provides 2 ways to define a route – @route and defroute. You can choose which to use.

@route is an annotation macro defined by using cl-annot. It takes a method, an URL-string and a function.

@route GET "/"
(defun index ()

;; A route with no name.
@route GET "/welcome"
(lambda (&key (|name| "Guest"))
  (format nil "Welcome, ~A" |name|))

This is similar to Caveman1's @url except its argument list. You don't have to specify an argument when you don't need it.

defroute is just a macro. It provides same feature to @route.

(defroute index "/" ()

;; A route with no name.
(defroute "/welcome" (&key (|name| "Guest"))
  (format nil "Welcome, ~A" |name|))

Since Caveman bases on ningle, Caveman also has the Sinatra-like routing system.

;; GET request (default)
@route GET "/" (lambda () ...)
(defroute ("/" :method :GET) () ...)

;; POST request
@route POST "/" (lambda () ...)
(defroute ("/" :method :POST) () ...)

;; PUT request
@route PUT "/" (lambda () ...)
(defroute ("/" :method :PUT) () ...)

;; DELETE request
@route DELETE "/" (lambda () ...)
(defroute ("/" :method :DELETE) () ...)

;; OPTIONS request
@route OPTIONS "/" (lambda () ...)
(defroute ("/" :method :OPTIONS) () ...)

;; For all methods
@route ANY "/" (lambda () ...)
(defroute ("/" :method :ANY) () ...)

Route pattern may contain “keyword” to put the value into the argument.

(defroute "/hello/:name" (&key name)
  (format nil "Hello, ~A" name))

The above controller will be invoked when you access to “/hello/Eitaro” or “/hello/Tomohiro”, and then name will be “Eitaro” and “Tomohiro”.

(&key name) is almost same as a lambda list of Common Lisp, excepts it always allows other keys.

(defroute "/hello/:name" (&rest params &key name)
  ;; ...

Route patterns may also contain “wildcard” parameters. They are accessible by splat.

(defroute "/say/*/to/*" (&key splat)
  ; matches /say/hello/to/world
  splat ;=> ("hello" "world")

(defroute "/download/*.*" (&key splat)
  ; matches /download/path/to/file.xml
  splat ;=> ("path/to/file" "xml")

If you'd like to write a regular expression for URL rule, :regexp t should work for it.

(defroute ("/hello/([\\w]+)" :regexp t) (&key captures)
  (format nil "Hello, ~A!" (first captures)))

Normally, routes are matched in the order they are defined. Only the first route matched is invoked and rest of them just will be ignored. But, a route can punt processing to the next matching route using next-route.

(defroute "/guess/:who" (&key who)
  (if (string= who "Eitaro")
      "You got me!"

(defroute "/guess/*" ()
  "You missed!")

You can return following formats as the result of defroute.

Structured query/post parameters

Parameter keys contain square brackets (“[“ & ”]”) will be parsed as structured parameters. You can access the parsed parameters as _parsed in routers.

<form action="/edit">
  <input type="name" name="person[name]" />
  <input type="name" name="person[email]" />
  <input type="name" name="person[birth][year]" />
  <input type="name" name="person[birth][month]" />
  <input type="name" name="person[birth][day]" />
(defroute "/edit" (&key _parsed)
  (format nil "~S" (cdr (assoc "person" _parsed :test #'string=))))
;=> "((\"name\" . \"Eitaro\") (\"email\" . \"\") (\"birth\" . ((\"year\" . 2000) (\"month\" . 1) (\"day\" . 1))))"

Blank keys mean they have multiple values.

<form action="/add">
  <input type="text" name="items[][name]" />
  <input type="text" name="items[][price]" />

  <input type="text" name="items[][name]" />
  <input type="text" name="items[][price]" />

  <input type="submit" value="Add" />
(defroute "/add" (&key _parsed)
  (format nil "~S" (assoc "items" _parsed :test #'string=)))
;=> "(((\"name\" . \"WiiU\") (\"price\" . \"30000\")) ((\"name\" . \"PS4\") (\"price\" . \"69000\")))"


Caveman adopts Djula for the default templating engine.

{% extends "layouts/default.html" %}
{% block title %}Users | MyApp{% endblock %}
{% block content %}
<div id="main">
  {% for user in users %}
    <li><a href="{{ user.url }}">{{ }}</a></li>
  {% endfor %}
{% endblock %}
(import 'myapp.view:render)

(render #P"users.html"
        '(:users ((:url "/id/1"
                   :name "nitro_idiot")
                  (:url "/id/2"
                   :name "meymao"))
          :has-next-page T))


This is an example of a JSON API.

(defroute "/user.json" (&key |id|)
  (let ((person (find-person-from-db |id|)))
    ;; person => (:|name| "Eitaro Fukamachi" :|email| "")
    (render-json person)))

;=> {"name":"Eitaro Fukamachi","email":""}

render-json is a part of a skeleton project. You can find its code in “src/view.lisp”.

Static file

Images, CSS, JS, favicon.ico and robot.txt in “static/” directory will be served by default.

/images/logo.png => {PROJECT_ROOT}/static/images/logo.png
/css/main.css    => {PROJECT_ROOT}/static/css/main.css
/js/app/index.js => {PROJECT_ROOT}/static/js/app/index.js
/robot.txt       => {PROJECT_ROOT}/static/robot.txt
/favicon.ico     => {PROJECT_ROOT}/static/favicon.ico

You can change these rules by rewriting “PROJECT_ROOT/app.lisp”. See Clack.Middleware.Static for detail.


Caveman adopts Envy as a configuration switcher. It allows to define multiple configurations and to switch them by an environment variable.

This is a typical example.

(defpackage :myapp.config
  (:use :cl
(in-package :myapp.config)

(setf (config-env-var) "APP_ENV")

(defconfig :common
  `(:application-root ,(asdf:component-pathname (asdf:find-system :myapp))))

(defconfig |development|
  '(:debug T
    ((:maindb :sqlite3 :database-name ,(merge-pathnames #P"test.db"

(defconfig |production|
    ((:maindb :mysql :database-name "myapp" :username "whoami" :password "1234")
     (:workerdb :mysql :database-name "jobs" :username "whoami" :password "1234"))))

(defconfig |staging|
  `(:debug T

Every configuration is a property list. You can choose the configuration which to use by setting APP_ENV.

To get a value from the current configuration, call myapp.config:config with a key you want.

(import 'myapp.config:config)

(setf (osicat:environment-variable "APP_ENV") "development")
(config :debug)
;=> T


When you add :databases to the configuration, Caveman enables database support. :databases is an association list of database settings.

(defconfig |production|
    ((:maindb :mysql :database-name "myapp" :username "whoami" :password "1234")
     (:workerdb :mysql :database-name "jobs" :username "whoami" :password "1234"))))

db in a package myapp.db is a function for connecting to each databases configured the above. Here is an example.

(use-package '(:myapp.db :sxql :datafly))

(defun search-adults ()
  (with-connection (db)
      (select :*
        (from :person)
        (where (:>= :age 20))))))

The connection is alive during the Lisp session and will be reused in each HTTP requests.

retrieve-all and the query language came from datafly and SxQL. See those documentations for more informations.

Set HTTP headers or HTTP status

There are several special variables available during a HTTP request. *request* and *response* represents a request and a response. If you are familiar with Clack, these are instances of subclasses of Clack.Request and Clack.Response.

(use-package :caveman2)

;; Get a value of Referer header.
(http-referer *request*)

;; Set Content-Type header.
(setf (getf (response-headers *response* :content-type) "application/json")

;; Set HTTP status.
(setf (status *response*) 304)

If you would like to set Content-Type “application/json” for all “*.json” requests, next-route will help you.

(defroute "/*.json" ()
  (setf (getf (response-headers *response*) :content-type) "application/json")

(defroute "/user.json" () ...)
(defroute "/search.json" () ...)
(defroute ("/new.json" :method :POST) () ...)

Using session

Session data is for memorizing user-specific data. *session* is a hash table represents session data.

This example increments :counter in the session and displays it for each visitors.

(defroute "/counter" ()
  (format nil "You came here ~A times."
          (incf (gethash :counter *session* 0))))

Caveman2 stores the session data in-memory by default. To change it, specify :store to :session in “PROJECT_ROOT/app.lisp”.

This example uses RDBMS to store it.

        :output (getf (config) :error-log))
- :session
+ (:session
+  :store (make-dbi-store :connector (lambda ()
+                                      (apply #'dbi:connect
+                                             (myapp.db:connection-settings)))))
  (if (productionp)
      (lambda (app)

NOTE: Don't forget to add :lack-session-store-dbi as :depends-on of your app. It is not a part of Clack/Lack.

See the source code of Lack.Session.Store.DBi for more informations.

Throw an HTTP status code

(import 'caveman2:throw-code)

(defroute ("/auth" :method :POST) (&key |name| |password|)
  (unless (authorize |name| |password|)
    (throw-code 403)))

Specify error pages

To specify error pages for 404, 500 or so, define a method on-exception of your app.

(defmethod on-exception ((app <web>) (code (eql 404)))
  (declare (ignore app code))
  (merge-pathnames #P"_errors/404.html"

Start a server

Your application has functions named start and stop to start/stop your web application. This is a example assuming that the name of your application is “myapp”.

(myapp:start :port 8080)

As Caveman bases on Clack/Lack, you can choose which server to run on – Hunchentoot, mod_lisp or FastCGI.

(myapp:start :server :hunchentoot :port 8080)
(myapp:start :server :fcgi :port 8080)

I recommend you to use Hunchentoot in local machine and use FastCGI for production environment.

You can also start your application by using Shelly.

$ APP_ENV=development shly -Lclack clackup app.lisp --server :fcgi --port 8080

Shelly allows you to execute a Common Lisp function like a shell command.

Hot Deployment

Though Caveman doesn't have a feature for hot deployment, Server::Starter – a Perl module – makes it easy.

$ APP_ENV=production start_server --port 8080 -- shly start --server :fcgi

To restart the server, send HUP signal (kill -HUP <pid>) to the start_server process.

Error Log

Caveman outputs error backtraces to a file which is specified at :error-log in your configuration.

(defconfig |default|
  `(:error-log #P"/var/log/apps/myapp_error.log"
    ((:maindb :sqlite3 :database-name ,(merge-pathnames #P"myapp.db"

Use another templating library


(import 'cl-who:with-html-output-to-string)

(defroute "/" ()
  (with-html-output-to-string (output nil :prologue t)
      (:head (:title "Welcome to Caveman!"))
      (:body "Blah blah blah."))))
;=> "<!DOCTYPE html PUBLIC \"-//W3C//DTD XHTML 1.0 Strict//EN\" \"\">
;   <html><head><title>Welcome to Caveman!</title></head><body>Blah blah blah.</body></html>"


(import 'cl-markup:xhtml)

(defroute "/" ()
    (:head (:title "Welcome to Caveman!"))
    (:body "Blah blah blah.")))
;=> "<?xml version=\"1.0\" encoding=\"UTF-8\"?><!DOCTYPE html PUBLIC \"-//W3C//DTD XHTML 1.0 Transitional//EN\" \"\"><html><head><title>Welcome to Caveman!</title></head><body>Blah blah blah.</body></html>"


{namespace myapp.view}

{template renderIndex}
<!DOCTYPE html>
  <title>"Welcome to Caveman!</title>
  Blah blah blah.
(import 'myapp.config:*template-directory*)

(closure-template:compile-cl-templates (merge-pathnames #P"index.tmpl"

(defroute "/" ()

See Also



Licensed under the LLGPL License.