clj-http documentation

Branches
Introduction
- Overview
- Philosophy
Installation
Quickstart
Caching
Authentication
Advanced Usage
Development
Tests
Testimonials
Other Libraries Providing Middleware
License

Branches

There are branches for the major version numbers:

2.x (no longer maintained except for security issues)
3.x (current stable releases and the main Github branch)
master (which is 4.x, unreleased, based on version 5 of the Apache HTTP Client)

Introduction

Overview

clj-http is an HTTP library wrapping the Apache HttpComponents client. This library has taken over from mmcgrana’s clj-http.

Philosophy

The design of clj-http is inspired by the Ring protocol for Clojure HTTP server applications.

The client in clj-http.core makes HTTP requests according to a given Ring request map and returns Ring response maps corresponding to the resulting HTTP response. The function clj-http.client/request uses Ring-style middleware to layer functionality over the core HTTP request/response implementation. Methods like clj-http.client/get are sugar over this clj-http.client/request function.

Installation

clj-http is available as a Maven artifact from Clojars.

With Leiningen/Boot:

[clj-http "3.13.0"]

If you need an older version, a 2.x release is also available.

[clj-http "2.3.0"]

clj-http 3.12.x supports clojure 1.6.0 and higher. clj-http 3.13.x supports clojure 1.8.0 and higher. clj-http 4.x will support clojure 1.7.0 and higher.

Quickstart

The main HTTP client functionality is provided by the clj-http.client namespace.

First, require it in the REPL:

(require '[clj-http.client :as client])

Or in your application:

(ns my-app.core
  (:require [clj-http.client :as client]))

The client supports simple get, head, put, post, delete, copy, move, patch, and options requests. Response are returned as Ring-style response maps:

HEAD

(client/head "http://example.com/resource")

(client/head "http://example.com/resource" {:accept :json})

GET

Example requests:

(client/get "http://example.com/resources/id")

;; Setting options
(client/get "http://example.com/resources/3" {:accept :json})
(client/get "http://example.com/resources/3" {:accept :json :query-params {"q" "foo, bar"}})

;; Specifying headers as either a string or collection:
(client/get "http://example.com"
            {:headers {"foo" ["bar" "baz"], "eggplant" "quux"}})

;; Using either string or keyword header names:
(client/get "http://example.com"
            {:headers {:foo ["bar" "baz"], :eggplant "quux"}})

;; Completely ignore cookies:
(client/post "http://example.com" {:cookie-policy :none})
;; There are also multiple ways to handle cookies
(client/post "http://example.com" {:cookie-policy :default})
(client/post "http://example.com" {:cookie-policy :netscape})
(client/post "http://example.com" {:cookie-policy :standard})
(client/post "http://example.com" {:cookie-policy :standard-strict})

;; Cookies can be completely configurable with a custom spec by adding a
;; function to return a cookie spec for parsing the cookie. For example, if you
;; wanted to configure a spec provider to have a certain compatibility level:
(client/post "http://example.com"
             {:cookie-spec
              (fn [http-context]
                (println "generating a new cookie spec")
                (.create
                 (org.apache.http.impl.cookie.RFC6265CookieSpecProvider.
                  org.apache.http.impl.cookie.RFC6265CookieSpecProvider$CompatibilityLevel/IE_MEDIUM_SECURITY
                  (PublicSuffixMatcherLoader/getDefault))
                 http-context))})
;; Or a version with relaxed compatibility
(client/post "http://example.com"
             {:cookie-spec
              (fn [http-context]
                (println "generating a new cookie spec")
                (.create
                 (org.apache.http.impl.cookie.RFC6265CookieSpecProvider.
                  org.apache.http.impl.cookie.RFC6265CookieSpecProvider$CompatibilityLevel/RELAXED
                  (PublicSuffixMatcherLoader/getDefault))
                 http-context))})

;; Sometimes you want to do your own validation or something, which you can do
;; by proxying the CookieSpecBase. Note that this doesn't actually return the
;; cookies, because clj-http does its own cookie parsing. If you want to store
;; the cookies from these methods you'll need to use a cookie store or put it in
;; some datastructure yourself.
(client/post "http://example.com"
             {:cookie-spec
              (fn [http-context]
                (proxy [org.apache.http.impl.cookie.CookieSpecBase] []
                  ;; Version and version header
                  (getVersion [] 0)
                  (getVersionHeader [] nil)
                  ;; parse headers into cookie objects
                  (parse [header cookie-origin] (java.util.ArrayList.))
                  ;; Validate a cookie, throwing MalformedCookieException if the
                  ;; cookies isn't valid
                  (validate [cookie cookie-origin]
                    (println "validating:" cookie))
                  ;; Determine if a cookie matches the target location
                  (match [cookie cookie-origin] true)
                  ;; Format a list of cookies into a list of headers
                  (formatCookies [cookies] (java.util.ArrayList.))))})

;; If you have created your own registry for cookie policies, you can provide
;; :cookie-policy-registry to use it. See
;; clj-http.core/create-custom-cookie-policy-registry for an example of a custom
;; registry
(client/post "http://example.com"
             {:cookie-policy-registry my-custom-policy-registry
              :cookie-policy "my-policy"})

;; Need to contact a server with an untrusted SSL cert?
(client/get "https://alioth.debian.org" {:insecure? true})

;; If you don't want to follow-redirects automatically:
(client/get "http://example.com/redirects-somewhere" {:redirect-strategy :none})

;; Only follow a certain number of redirects:
(client/get "http://example.com/redirects-somewhere" {:max-redirects 5})

;; Avoid throwing exceptions if redirected too many times:
(client/get "http://example.com/redirects-somewhere" {:max-redirects 5 :redirect-strategy :graceful})

;; Throw an exception if the get takes too long. Timeouts in milliseconds.
(client/get "http://example.com/redirects-somewhere" {:socket-timeout 1000 :connection-timeout 1000})

;; Query parameters
(client/get "http://example.com/search" {:query-params {"q" "foo, bar"}})

;; "Nested" query parameters
;; (this yields a query string of `a[e][f]=6&a[b][c]=5`)
(client/get "http://example.com/search" {:query-params {:a {:b {:c 5} :e {:f 6}}}})

;; Provide cookies — uses same schema as :cookies returned in responses
;; (see the cookie store option for easy cross-request maintenance of cookies)
(client/get "http://example.com"
            {:cookies {"ring-session" {:discard true, :path "/", :value "", :version 0}}})

;; Tell clj-http not to decode cookies from the response header
(client/get "http://example.com" {:decode-cookies false})

;; Support for IPv6!
(client/get "http://[2001:62f5:9006:e472:cabd:c8ff:fee3:8ddf]")

;; Super advanced, your own http-client-context and request-config
(client/get "http://example.com/get"
            {:http-client-context my-http-client-context
             :http-request-config my-request-config})

The client will also follow redirects on the appropriate 30* status codes.

The client transparently accepts and decompresses the gzip and deflate content encodings.

:trace-redirects will contain the chain of the redirections followed.

PUT

(client/put "http://example.com/api" {:body "my PUT body"})

POST

;; Various options:
(client/post "http://example.com/api"
             {:basic-auth ["user" "pass"]
              :body "{\"json\": \"input\"}"
              :headers {"X-Api-Version" "2"}
              :content-type :json
              :socket-timeout 1000      ;; in milliseconds
              :connection-timeout 1000  ;; in milliseconds
              :accept :json})

;; Send form params as a urlencoded body (POST or PUT)
(client/post "http://example.com" {:form-params {:foo "bar"}})

;; Send form params as a json encoded body (POST or PUT)
(client/post "http://example.com" {:form-params {:foo "bar"} :content-type :json})

;; Send form params as a json encoded body (POST or PUT) with options
(client/post "http://example.com" {:form-params {:foo "bar"}
                                   :content-type :json
                                   :json-opts {:date-format "yyyy-MM-dd"}})

;; You can also specify the encoding of form parameters
(client/post "http://example.com" {:form-params {:foo "bar"}
                                   :form-param-encoding "ISO-8859-1"})

;; Send form params as a Transit encoded JSON body (POST or PUT) with options
(client/post "http://example.com" {:form-params {:foo "bar"}
                                   :content-type :transit+json
                                   :transit-opts
                                   {:encode {:handlers {}}
                                    :decode {:handlers {}}}})

;; Send form params as a Transit encoded MessagePack body (POST or PUT) with options
(client/post "http://example.com" {:form-params {:foo "bar"}
                                   :content-type :transit+msgpack
                                   :transit-opts
                                   {:encode {:handlers {}}
                                    :decode {:handlers {}}}})

;; Multipart form uploads/posts
;; takes a vector of maps, to preserve the order of entities, :name
;; will be used as the part name unless :part-name is specified
(client/post "http://example.org" {:multipart [{:name "title" :content "My Awesome Picture"}
                                               {:name "Content/type" :content "image/jpeg"}
                                               {:name "foo.txt" :part-name "eggplant" :content "Eggplants"}
                                               {:name "file" :content (clojure.java.io/file "pic.jpg")}]
                                   ;; You can also optionally pass a :mime-subtype
                                   :mime-subtype "foo"})

;; Multipart :content values can be one of the following:
;; String, InputStream, File, a byte-array, or an instance of org.apache.http.entity.mime.content.ContentBody
;; Some Multipart bodies can also support more keys (like :encoding
;; and :mime-type), check src/clj-http/multipart.clj to see all flags

;; Apache's http client automatically retries on IOExceptions, if you
;; would like to handle these retries yourself, you can specify a
;; :retry-handler. Return true to retry, false to stop trying:
(client/post "http://example.org" {:multipart [["title" "Foo"]
                                               ["Content/type" "text/plain"]
                                               ["file" (clojure.java.io/file "/tmp/missing-file")]]
                                   :retry-handler (fn [ex try-count http-context]
                                                    (println "Got:" ex)
                                                    (if (> try-count 4) false true))})

;; to handle a file with non-ascii filename, try :multipart-charset "UTF-8" and :multipart-mode BROWSER_COMPATIBLE
;; see also: https://stackoverflow.com/questions/3393445/international-characters-in-filename-in-mutipart-formdata
(import (org.apache.http.entity.mime HttpMultipartMode))

(client/post "http://example.org" {:multipart [{:content (clojure.java.io/file "日本語.txt")}]
                                   :multipart-mode HttpMultipartMode/BROWSER_COMPATIBLE
                                   :multipart-charset "UTF-8"} )

A word about flattening nested :query-params and :form-params maps. There are essentially three different ways to handle flattening them:

:ignore-nested-query-string: Do not handle nested query parameters specially, treat them as the exact text they come in as. Defaults to false.
:flatten-nested-form-params: Flatten nested (map within a map) :form-params before encoding it as the body. Defaults to false, meaning form params are encoded only x-www-form-urlencoded.
:flatten-nested-keys: An advanced way of specifying which keys having nested maps should be flattened. A middleware function checks the previous two options (:ignore-nested-query-string and :flatten-nested-form-params) and modifies this to be the list that will be flattened.

DELETE

(client/delete "http://example.com/resource")

Async HTTP Request

The new async HTTP request API is a Ring-style async API. All options for synchronous requests can be used in asynchronous requests. Starting an async request is easy, for example:

;; :async? in options map need to be true
(client/get "http://example.com"
            {:async? true}
            ;; respond callback
            (fn [response] (println "response is:" response))
            ;; raise callback
            (fn [exception] (println "exception message is: " (.getMessage exception))))

All exceptions thrown during the request will be passed to the raise callback.

Cancelling Requests

Calls to the http methods with :async true return an Apache BasicFuture that you can call .get or .cancel on. See the Javadocs for BasicFuture here. For instance:

(import '(java.util.concurrent TimeoutException TimeUnit))

(let [future (client/get "http://example.com/slow-url"
                         {:async true :oncancel #(println "request was cancelled")}
                         #(println :got %) #(println :err %))]
  (try
    (.get future 1 TimeUnit/SECONDS)
    (catch TimeoutException e
      ;; Cancel the request, it's taken too long
      (.cancel future true))))

Coercions

clj-http allows coercing the body of the request either before it is sent (input coercion), or after it’s received (output coercion) from the server.

Input coercion

;; body as a byte-array
(client/post "http://example.com/resources" {:body my-byte-array})

;; body as a string
(client/post "http://example.com/resources" {:body "string"})

;; :body-encoding is optional and defaults to "UTF-8"
(client/post "http://example.com/resources"
             {:body "string" :body-encoding "UTF-8"})

;; body as a file
(client/post "http://example.com/resources"
             {:body (clojure.java.io/file "/tmp/foo") :body-encoding "UTF-8"})

;; :length is optional for passing in an InputStream; if not
;; supplied it will default to -1 to signal to HttpClient to use
;; chunked encoding
(client/post "http://example.com/resources"
             {:body (clojure.java.io/input-stream "/tmp/foo")})

(client/post "http://example.com/resources"
             {:body (clojure.java.io/input-stream "/tmp/foo") :length 1000})

Output coercion

;; The default output is a string body
(client/get "http://example.com/foo.txt")

;; Coerce as a byte-array
(client/get "http://example.com/favicon.ico" {:as :byte-array})

;; Coerce as something other than UTF-8 string
(client/get "http://example.com/string.txt" {:as "UTF-16"})

;; Coerce as json
(client/get "http://example.com/foo.json" {:as :json})
(client/get "http://example.com/foo.json" {:as :json-string-keys})

;; Coerce as Transit encoded JSON or MessagePack
(client/get "http://example.com/foo" {:as :transit+json})
(client/get "http://example.com/foo" {:as :transit+msgpack})

;; Coerce as a clojure datastructure
(client/get "http://example.com/foo.clj" {:as :clojure})

;; Coerce as x-www-form-urlencoded
(client/post "http://example.com/foo" {:as :x-www-form-urlencoded})

;; Try to automatically coerce the output based on the content-type
;; header (this is currently a BETA feature!). Currently supports
;; text, json and clojure (with automatic charset detection)
;; clojure coercion requires "application/clojure" or
;; "application/edn" in the content-type header
(client/get "http://example.com/foo.json" {:as :auto})

;; Return the body as a stream
(client/get "http://example.com/bigrequest.html" {:as :stream})
;; Note that the connection to the server will NOT be closed until the
;; stream has been read

;; Return the body as a java.io.BufferedReader
(client/get "http://example.com/bigrequest.html" {:as :reader})
;; As above, the connection will remain open until the stream has been
;; read.  The reader will attempt to respect the server-specified charset,
;; if any, defaulting to UTF-8.

Output coercion with :as :json, :as :json-string-keys or :as :x-www-form-urlencoded, will only work with an optional dependency, see Optional Dependencies.

By default, JSON coercion is only applied when the response’s status is considered “unexceptional”. If the :unexceptional-status option is provided, then its value is a function which specifies what status codes are unexceptional. :unexceptional-status defaults to clj-http.client/unexceptional-status?.

If you would like to change under what conditions coercion is applied, you can send the :coerce option, which can be set to:

:always        ;; always json decode the body
:unexceptional ;; json decode when an HTTP response is considered unexceptional
:exceptional   ;; json decode when an HTTP response is considered exceptional

The :coerce setting defaults to :unexceptional.

Headers

clj-http’s treatment of headers is a little more permissive than the ring spec specifies.

Rather than forcing all request headers to be lowercase strings, clj-http allows strings or keywords of any case. Keywords will be transformed into their canonical representation, so the :content-md5 header will be sent to the server as “Content-MD5”, for instance. String keys in request headers, however, will be sent to the server with their casing unchanged.

Response headers can be read as keywords or strings of any case. If the server responds with a “Date” header, you could access the value of that header as :date, “date”, “Date”, etc.

If for some reason you require access to the original header name that the server specified, it is available by invoking (keys …) on the header map.

This special treatment of headers is implemented in the wrap-header-map middleware, which (like any middleware) can be disabled by using with-middleware to specify different behavior.

Query-string parameters

There are four different ways that query string parameters for array values can be generated, depending on what the resulting query string should look like, they are:

A repeating parameter (default)
Array style
Indexed array style
Comma separated style

Here is an example of the input and output for the :query-params parameter, controlled by the :multi-param-style option:

;; default style, with :multi-param-style unset
:a [1 2 3] => "a=1&a=2&a=3"
;; with :multi-param-style :array, a repeating param with array suffix
;; (PHP-style):
:a [1 2 3] => "a[]=1&a[]=2&a[]=3"
;; with :multi-param-style :indexed, a repeating param with array suffix and
;; index (Rails-style):
:a [1 2 3] => "a[0]=1&a[1]=2&a[2]=3"
;; with :multi-param-style :comma-separated, a param with comma-separated values
:a [1 2 3] => "a=1,2,3"

Meta Tag Headers

HTML 4.01 allows using the tag <meta http-equiv="..." /> and HTML 5 allows using the tag <meta charset="..." /> to specify a header that should be treated as an HTTP response header. By default, clj-http will ignore the body of the response (other than the regular output coercion), but if you need clj-http to parse the headers out of the body, you can use the :decode-body-headers option:

;; without decoding body headers (defaults to off):
(:headers (client/get "http://www.yomiuri.co.jp/"))
=> {"server" "Apache",
    "content-encoding" "gzip",
    "content-type" "text/html",
    "date" "Tue, 09 Oct 2012 18:02:41 GMT",
    "cache-control" "max-age=0, no-cache",
    "expires" "Tue, 09 Oct 2012 18:02:41 GMT",
    "etag" "\"1dfb-2686-4cba2686fb8b1\"",
    "pragma" "no-cache",
    "connection" "close"}

;; with decoding body headers, notice the content-type,
;; content-style-type and content-script-type headers:
(:headers (client/get "http://www.yomiuri.co.jp/" {:decode-body-headers true}))
=> {"server" "Apache",
    "content-encoding" "gzip",
    "content-script-type" "text/javascript",
    "content-style-type" "text/css",
    "content-type" "text/html; charset=Shift_JIS",
    "date" "Tue, 09 Oct 2012 18:02:59 GMT",
    "cache-control" "max-age=0, no-cache",
    "expires" "Tue, 09 Oct 2012 18:02:59 GMT",
    "etag" "\"1dfb-2686-4cba2686fb8b1\"",
    "pragma" "no-cache",
    "connection" "close"}

This can be used to have clj-http correctly interpret the body’s charset by using:

(client/get "http://www.yomiuri.co.jp/" {:decode-body-headers true :as :auto})
=> ;; correctly formatted :body (Shift_JIS charset instead of UTF-8)

Note that this feature is currently beta and uses Crouton to parse the body of the request. If you want to use this feature, you can include Crouton in addition to clj-http as a dependency like so:

(defproject foo "0.1.0-SNAPSHOT"
  :dependencies [[org.clojure/clojure "1.3.0"]
                 [clj-http "0.6.0"]
                 [crouton "1.0.0"]])

Note also that HEAD requests will not return a body, in which case this setting will have no effect.

clj-http will automatically disable the :decode-body-headers option.

Link Headers

clj-http parses any link headers returned in the response, and adds them to the :links key on the response map. This is particularly useful for paging RESTful APIs:

(:links (client/get "https://api.github.com/gists"))
=> {:next {:href "https://api.github.com/gists?page=2"}
    :last {:href "https://api.github.com/gists?page=22884"}}

Redirects

clj-http conforms its behaviour regarding automatic redirects to the RFC.

It means that redirects on status 301, 302, 307 and 308 are not redirected on methods other than GET and HEAD. If you want a behaviour closer to what most browser have, you can set :redirect-strategy to :lax in your request to have automatic redirection work on all methods by transforming the method of the request to GET.

Redirect Options:

:trace-redirects

If true, clj-http will enhance the response object with a list of redirected URLs with key: :trace-redirects.

:redirect-strategy

Sets the redirect strategy for clj-http. Accepts the following:

:none - Perform no redirects
:default - See https://hc.apache.org/httpcomponents-client-ga/httpclient/apidocs/org/apache/http/impl/client/DefaultRedirectStrategy.html
:lax - See https://hc.apache.org/httpcomponents-client-ga/httpclient/apidocs/org/apache/http/impl/client/LaxRedirectStrategy.html
:graceful - Similar to :default, but does not throw exceptions when max redirects is reached. This is the redirects behaviour in 2.x
nil - When nil, assumes :default

You may also pass in an instance of RedirectStrategy (in the :redirect-strategy key) if you want a behavior that’s not implemented.

Additionally, clj-http will attempt to validate that a redirect host is not invalid, you can disable this by setting :validate-redirects false in the request (the default is true)

NOTE: The options :force-redirects and :follow-redirects (present in clj-http 2.x are no longer used). You can use :graceful to mostly emulate the old redirect behaviour.

How to create a custom RedirectStrategy

As mentioned earlier, it’s possible to pass a custom instance of RedirectStrategy. The snippet below shows how to create a custom RedirectStrategy by wrapping the default strategy.

(def default-strategy org.apache.http.impl.client.DefaultRedirectStrategy/INSTANCE)

(def logging-redirect-strategy
  (reify org.apache.http.client.RedirectStrategy
    (getRedirect [this request response context]
      (println "attempting redirect...")
      (.getRedirect default-strategy request response context))
    (isRedirected [this request response context]
      (println "checking isRedirected")
      (.isRedirected default-strategy request response context))))

(client/get "https://httpbin.org/absolute-redirect/3" {:redirect-strategy logging-redirect-strategy})
;; this will output the following:
;;
;;   checking isRedirected
;;   attempting redirect...
;;   checking isRedirected
;;   attempting redirect...
;;   checking isRedirected
;;   attempting redirect...
;;   checking isRedirected

Cookies

Cookiestores

clj-http can simplify the maintenance of cookies across requests if it is provided with a cookie store.

(binding [clj-http.core/*cookie-store* (clj-http.cookies/cookie-store)]
  (client/post "http://example.com/login" {:form-params {:username "..."
                                                         :password "..."}})
  (client/get "http://example.com/secured-page")
  ...)

(The clj-http.cookies/cookie-store function returns a new empty instance of a default implementation of org.apache.http.client.CookieStore.)

This will allow cookies to only be written to the cookie store. Cookies from the cookie-store will not automatically be sent with future requests.

If you would like cookies from the cookie-store to automatically be sent with each request, specify the cookie-store with the :cookie-store option:

(let [my-cs (clj-http.cookies/cookie-store)]
  (client/post "http://example.com/login" {:form-params {:username "..."
                                                         :password "..."}
                                           :cookie-store my-cs})
  (client/post "http://example.com/update" {:body my-data
                                            :cookie-store my-cs}))

You can also use the get-cookies function to retrieve the cookies from a cookie store:

(def cs (clj-http.cookies/cookie-store))

(client/get "http://google.com" {:cookie-store cs})

(clojure.pprint/pprint (clj-http.cookies/get-cookies cs))
{"NID"
 {:domain ".google.com",
  :expires #<Date Tue Oct 02 10:12:06 MDT 2012>,
  :path "/",
  :value
  "58=c387....",
  :version 0},
 "PREF"
 {:domain ".google.com",
  :expires #<Date Wed Apr 02 10:12:06 MDT 2014>,
  :path "/",
  :value
  "ID=3ba...:FF=0:TM=133...:LM=133...:S=_iRM...",
  :version 0}}

Keystores, Trust-stores

You can also specify your own keystore/trust-store to be used:

(client/get "https://example.com" {:keystore "/path/to/keystore.ks"
                                   :keystore-type "jks" ; default: jks
                                   :keystore-pass "secretpass"
                                   :trust-store "/path/to/trust-store.ks"
                                   :trust-store-type "jks" ; default jks
                                   :trust-store-pass "trustpass"})

The :keystore/:trust-store values may be either paths to keystore files or KeyStore instances.

Exceptions

The client will throw exceptions on, well, exceptional status codes, meaning all HTTP responses other than #{200 201 202 203 204 205 206 207 300 301 302 303 304 307}. clj-http will throw a Slingshot Stone that can be caught by a regular (catch Exception e ...) or in Slingshot’s try+ block:

(client/get "http://example.com/broken")
=> ExceptionInfo clj-http: status 404  clj-http.client/wrap-exceptions/fn--583 (client.clj:41)
;; Or, if you would like the Exception message to contain the entire response:
(client/get "http://example.com/broken" {:throw-entire-message? true})
=> ExceptionInfo clj-http: status 404 {:status 404,
                                       :headers {"server" "nginx/1.0.4",
                                                 "x-runtime" "12ms",
                                                 "content-encoding" "gzip",
                                                 "content-type" "text/html; charset=utf-8",
                                                 "date" "Mon, 17 Oct 2011 23:15 :36 GMT",
                                                 "cache-control" "no-cache",
                                                 "status" "404 Not Found",
                                                 "transfer-encoding" "chunked",
                                                 "connection" "close"},
                                       :body "...body here..."}
   clj-http.client/wrap-exceptions/fn--584 (client.clj:42

;; You can also ignore HTTP-status-code exceptions and handle them yourself:
(client/get "http://example.com/broken" {:throw-exceptions false})
;; Or ignore an unknown host (methods return 'nil' if this is set to
;; true and the host does not exist:
(client/get "http://example.invalid" {:ignore-unknown-host? true})
;; Or customize the http statuses that will not throw:
(client/get "http://example.com/broken" {:unexceptional-status #(<= 200 % 299)})

(spacing added by me to be human readable)

How to use with Slingshot:

; Response map is thrown as exception obj.
; We filter out by status codes
(try+
  (client/get "http://example.com/broken")
  (catch [:status 403] {:keys [request-time headers body]}
    (log/warn "403" request-time headers))
  (catch [:status 404] {:keys [request-time headers body]}
    (log/warn "NOT Found 404" request-time headers body))
  (catch Object _
    (log/error (:throwable &throw-context) "unexpected error")
    (throw+)))

Decompression

By default, clj-http will add the {"Accept-Encoding" "gzip, deflate"} header to requests, and automatically decompress the resulting gzip or deflate stream if the Content-Encoding header is found on the response. If this is undesired, the {:decompress-body false} option can be specified:

;; Auto-decompression used: (google requires a user-agent to send gzip data)
(def h {"User-Agent" "Mozilla/5.0 (Windows NT 6.1;) Gecko/20100101 Firefox/13.0.1"})
(def resp (client/get "http://google.com" {:headers h}))
(:orig-content-encoding resp)
=> "gzip" ;; <= google sent response gzipped

;; and without decompression:
(def resp2 (client/get "http://google.com" {:headers h :decompress-body false})
(:orig-content-encoding resp2)
=> nil

If clj-http decompresses something, the “content-encoding” header is removed from the headers (because the encoding is no longer true). This allows clj-http to be used as a pass-through proxy with ring. The original content-encoding is available as :orig-content-encoding in the response map if auto-decompression is enabled.

Debugging

There are four debugging methods you can use:

;; print request info to *out*:
(client/get "http://example.org" {:debug true})

;; print request info to *out*, including request body:
(client/post "http://example.org" {:debug true :debug-body true :body "..."})

;; save the request that was sent in a :request key in the response:
(client/get "http://example.org" {:save-request? true})

;; save the request that was sent in a :request key in the response,
;; including the body content:
(client/get "http://example.org" {:save-request? true :debug-body true})

;; add an HttpResponseInterceptor to the request. This callback
;; is called for each redirects with the following args:
;; ^HttpResponse resp, HttpContext^ ctx
;; this allows low level debugging + access to socket.
;; see http://hc.apache.org/httpcomponents-core-ga/httpcore/apidocs/org/apache/http/HttpResponseInterceptor.html
(client/get "http://example.org" {:response-interceptor (fn [resp ctx] (println ctx))})

Logging

Finally, if you want to access the logging that the Apache client does internally, you can set up your dependencies to add the log4j2 libraries and configure the logging for clj-http. In order to do this, you’ll need to add

[org.apache.logging.log4j/log4j-api "2.11.0"]
[org.apache.logging.log4j/log4j-core "2.11.0"]
[org.apache.logging.log4j/log4j-1.2-api "2.11.0"]

To your project.clj and have a usable log4j2.properties. I have provided one in resources/log4j2.properties. Make sure to set:

rootLogger.level = debug

If you want to see debug information (or “trace” for trace logging). When you perform a request you should see something akin to this in the logs:

[2018-03-20T20:36:34,635][DEBUG][o.a.h.c.p.RequestAddCookies] CookieSpec selected: default
[2018-03-20T20:36:34,635][DEBUG][o.a.h.c.p.RequestAuthCache] Auth cache not set in the context
[2018-03-20T20:36:34,635][DEBUG][o.a.h.i.c.BasicHttpClientConnectionManager] Get connection for route {s}->https://example.com:443
[2018-03-20T20:36:34,636][DEBUG][o.a.h.i.c.DefaultManagedHttpClientConnection] http-outgoing-1: set socket timeout to 0
[2018-03-20T20:36:34,636][DEBUG][o.a.h.i.e.MainClientExec ] Opening connection {s}->https://example.com:443
[2018-03-20T20:36:34,644][DEBUG][o.a.h.i.c.DefaultHttpClientConnectionOperator] Connecting to example.com/10.0.0.1:443
[2018-03-20T20:36:34,644][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory] Connecting socket to example.com/10.0.0.1:443 with timeout 0
[2018-03-20T20:36:34,692][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory] Enabled protocols: [TLSv1, TLSv1.1, TLSv1.2]
[2018-03-20T20:36:34,693][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory] Enabled cipher suites:[TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384, ... etc ...]
[2018-03-20T20:36:34,693][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory] Starting handshake
[2018-03-20T20:36:34,841][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory] Secure session established
[2018-03-20T20:36:34,842][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory]  negotiated protocol: TLSv1.2
[2018-03-20T20:36:34,842][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory]  negotiated cipher suite: TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
[2018-03-20T20:36:34,843][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory]  peer principal: CN=example.com
[2018-03-20T20:36:34,843][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory]  peer alternative names: [example.com, www.example.com]
[2018-03-20T20:36:34,843][DEBUG][o.a.h.c.s.SSLConnectionSocketFactory]  issuer principal: CN=Let's Encrypt Authority X3, O=Let's Encrypt, C=US
[2018-03-20T20:36:34,844][DEBUG][o.a.h.i.c.DefaultHttpClientConnectionOperator] Connection established 192.168.0.29:36792<->10.0.0.1:443
[2018-03-20T20:36:34,844][DEBUG][o.a.h.i.e.MainClientExec ] Executing request POST /post HTTP/1.1
[2018-03-20T20:36:34,844][DEBUG][o.a.h.i.e.MainClientExec ] Target auth state: UNCHALLENGED
[2018-03-20T20:36:34,844][DEBUG][o.a.h.i.e.MainClientExec ] Proxy auth state: UNCHALLENGED
[2018-03-20T20:36:34,845][DEBUG][o.a.h.headers            ] http-outgoing-1 >> POST /post HTTP/1.1
[2018-03-20T20:36:34,845][DEBUG][o.a.h.headers            ] http-outgoing-1 >> Connection: close
[2018-03-20T20:36:34,845][DEBUG][o.a.h.headers            ] http-outgoing-1 >> accept-encoding: gzip, deflate
[2018-03-20T20:36:34,845][DEBUG][o.a.h.headers            ] http-outgoing-1 >> Content-Length: 14
[2018-03-20T20:36:34,845][DEBUG][o.a.h.headers            ] http-outgoing-1 >> Content-Type: text/plain; charset=UTF-8
[2018-03-20T20:36:34,846][DEBUG][o.a.h.headers            ] http-outgoing-1 >> Host: example.com
[2018-03-20T20:36:34,846][DEBUG][o.a.h.headers            ] http-outgoing-1 >> User-Agent: Apache-HttpClient/4.5.5 (Java/9.0.1)
[2018-03-20T20:36:34,846][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "POST /post HTTP/1.1[\r][\n]"
[2018-03-20T20:36:34,846][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "Connection: close[\r][\n]"
[2018-03-20T20:36:34,846][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "accept-encoding: gzip, deflate[\r][\n]"
[2018-03-20T20:36:34,847][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "Content-Length: 14[\r][\n]"
[2018-03-20T20:36:34,847][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "Content-Type: text/plain; charset=UTF-8[\r][\n]"
[2018-03-20T20:36:34,847][DEBUG][o.a.h.wire               ] http-outgoing-1 >> "Host: example.com[\r][\n]"
etc etc it will go on forever and be very verbose

This provides both the data sent and received on the wire for debugging purposes.

I’ve also provided an example for changing the log level from clojure in examples/logging-apache-requests.clj.

Caching

clj-http supports Apache’s caching client, essentially it “provides an HTTP/1.1-compliant caching layer to be used with HttpClient–the Java equivalent of a browser cache.” (see the explanation in the apache docs). In order to use the cache, a reusable connection manager and http-client must be used.

An example of basic usage with the default options:

(let [cm (conn/make-reusable-conn-manager {})
      client (:http-client (http/get "http://example.com"
                                     {:connection-manager cm :cache true}))]
  (http/get "http://example.com"
            {:connection-manager cm :http-client client :cache true})
  (http/get "http://example.com"
            {:connection-manager cm :http-client client :cache true})
  (http/get "http://example.com"
            {:connection-manager cm :http-client client :cache true}))

You can build your own cache config by providing either a map of caching configuration options, or by providing a CacheConfig object, as seen below:

(let [cm (conn/make-reusable-conn-manager {})
      cache-config (core/build-cache-config
                    {:cache-config {:max-object-size 4096}})
      client (:http-client (http/get "http://example.com"
                                     {:connection-manager cm :cache true}))]
  (http/get "http://example.com"
            ;; Use the default cache config settings
            {:connection-manager cm :http-client client :cache true})
  (http/get "http://example.com"
            {:connection-manager cm :http-client client :cache true
             ;; Provide cache configuration options as a map
             :cache-config {:max-object-size 9152
                            :max-cache-entries 100}})
  (http/get "http://example.com"
            {:connection-manager cm :http-client client :cache true
             ;; Provide the cache configuration as a CacheConfig object
             :cache-config cache-config}))

In the response, clj-http provides the :cached key to indicate whether the response was cached, missed, etc:

nil: Caching was not used for this request
:CACHE_HIT: A response was generated from the cache with no requests sent upstream.
:CACHE_MISS: The response came from an upstream server.
:CACHE_MODULE_RESPONSE: The response was generated directly by the caching module.
:VALIDATED: The response was generated from the cache after validating the entry with the origin server.

Authentication

Basic Auth

(client/get "http://example.com/protected" {:basic-auth ["user" "pass"]})
(client/get "http://example.com/protected" {:basic-auth "user:pass"})

Digest Auth

(client/get "http://example.com/protected" {:digest-auth ["user" "pass"]})

NTLM Auth

(client/get "http://example.com/protected" {:ntlm-auth ["user" "pass" "host" "domain"]})

oAuth2

(client/get "http://example.com/protected" {:oauth-token "secret-token"})

Advanced Usage

Raw Request

A more general request function is also available, which is useful as a primitive for building higher-level interfaces:

(defn api-action [method path & [opts]]
  (client/request
    (merge {:method method :url (str "http://example.com/" path)} opts)))

Boolean options

Since 0.9.0, all boolean options can be expressed as either {:debug true} or {:debug? true}, with or without the question mark.

Persistent Connections

clj-http can use persistent connections to speed up connections if multiple connections are being used:

(with-connection-pool {:timeout 5 :threads 4 :insecure? false :default-per-route 10}
  (get "http://example.org/1")
  (post "http://example.org/2")
  (get "http://example.org/3")
  ...
  (get "http://example.org/999"))

For async request, you can use with-async-connection-pool

(with-async-connection-pool {:timeout 5 :threads 4 :insecure? false :default-per-route 10}
  (get "http://example.org/1" {:async? true} resp1 exce1)
  (post "http://example.org/2" {:async? true} resp2 exce2)
  (get "http://example.org/3" {:async? true} resp3 exce3)
  ...
  (get "http://example.org/999" {:async? true} resp999 exce999))

This is MUCH faster than sequentially performing all requests, because a persistent connection can be used instead creating a new connection for each request.

If you want to start an async request in the respond callback of an async request and reuse the pool context, just use reuse-pool.

(with-async-connection-pool {:timeout 5 :threads 4 :insecure? false :default-per-route 10}
  (get "http://example.org/1" {:async? true} resp1 exce1)
  (post "http://example.org/2"
        {:async? true}
        (fn [resp] (get "http://example.org/3"
                        (reuse-pool {:async? true} resp)
                        resp3 exce3))
        exce2))

There are many advanced options available when creating asynchronous connection pools that can be configured by passing an :io-config map in the connection manager parameters. It supports:

:connect-timeout
:interest-op-queued
:io-thread-count
:rcv-buf-size
:select-interval
:shutdown-grace-period
:snd-buf-size
:so-keep-alive
:so-linger
:so-timeout
:tcp-no-delay

See the docstring on with-async-connection-pool for more information about these options.

If you would prefer to handle managing the connection manager yourself, you can create a connection manager and specify it for each request:

(def cm (clj-http.conn-mgr/make-reusable-conn-manager {:timeout 2 :threads 3}))
(def cm2 (clj-http.conn-mgr/make-reusable-conn-manager {:timeout 10 :threads 1}))

(get "http://example.org/1" {:connection-manager cm2})
(post "http://example.org/2" {:connection-manager cm})
(get "http://example.org/3" {:connection-manager cm2})

;; Don't forget to shut it down when you're done!
(clj-http.conn-mgr/shutdown-manager cm)
(clj-http.conn-mgr/shutdown-manager cm2)

See the docstring on make-reusable-conn-manager for options and default values.

In the current version, pooled async request CANNOT specify connection manager.

Re-using `HttpClient` between requests

In some cases, you may want to re-use the same HttpClient object between requests, either so you don’t have to build it every time, or because you make some configuration change to the request. clj-http will return the built HTTP client in :http-client which you can then specify in subsequent requests (with :http-client). Note that in order to reuse the client a connection manager must be used.

;; Re-use the HttpClient clj-http builds for you:
(let [cm (conn/make-reusable-conn-manager {})
      resp (client/get "http://example.com" {:connection-manager cm})
      hclient (:http-client resp)]
  (client/get "http://example.com/1"
              {:connection-manager cm :http-client hclient})
  (client/get "http://example.com/2"
              {:connection-manager cm :http-client hclient})
  (client/get "http://example.com/3"
              {:connection-manager cm :http-client hclient}))

;; You can also build your own, using clj-http's helper or manually building it:
(let [cm (conn/make-reusable-conn-manager {})
      hclient (core/build-http-client {} false cm)]
  (client/get "http://example.com/1"
              {:connection-manager cm :http-client hclient})
  (client/get "http://example.com/2"
              {:connection-manager cm :http-client hclient})
  (client/get "http://example.com/3"
              {:connection-manager cm :http-client hclient}))

;; Async http clients may also be created and re-used:
(let [acm (conn/make-reuseable-async-conn-manager {})
      ahclient (core/build-async-http-client {} acm)]
  (client/get "http://example.com/1"
              {:connection-manager cm :http-client ahclient}
              handle-response handle-failure)
  (client/get "http://example.com/2"
              {:connection-manager cm :http-client ahclient}
              handle-response handle-failure)
  (client/get "http://example.com/3"
              {:connection-manager cm :http-client ahclient}
              handle-response handle-failure))

Proxies

A proxy can be specified by setting the Java properties: <scheme>.proxyHost and <scheme>.proxyPort where <scheme> is the client scheme used (normally ‘http’ or ‘https’). http.nonProxyHosts allows you to specify a pattern for hostnames which do not require proxy routing - this is shared for all schemes. Additionally, per-request proxies can be specified with the proxy-host and proxy-port options (this overrides http.nonProxyHosts too):

(client/get "http://example.com" {:proxy-host "127.0.0.1" :proxy-port 8118})

Proxy credentials can also be explicitly set as

(client/get "http://example.com" {:proxy-host "127.0.0.1" :proxy-port 8118 :proxy-user "proxy-user" :proxy-pass "superSecurePassword"})

You can also specify the proxy-ignore-hosts parameter with a list of hosts where the proxy should be ignored. By default this list is #{"localhost" "127.0.0.1"}.

A SOCKS proxy can be used by creating a proxied connection manager with clj-http.conn-mgr/make-socks-proxied-conn-manager. Then using that connection manager in the request.

For example if you wanted to connect to a local socks proxy on port 8081 you would:

(ns foo.bar
  (:require [clj-http.client :as client]
            [clj-http.conn-mgr :as conn-mgr]))

(client/get "https://google.com"
            {:connection-manager
             (conn-mgr/make-socks-proxied-conn-manager "localhost" 8081)})

If your SOCKS connection requires a keystore / trust-store, you can specify that too:

(ns foo.bar
  (:require [clj-http.client :as client]
            [clj-http.conn-mgr :as conn-mgr]))

(client/get "https://google.com"
            {:connection-manager
             (conn-mgr/make-socks-proxied-conn-manager "localhost" 8081
               {:keystore "/path/to/keystore.ks"
                :keystore-type "jks" ; default: jks
                :keystore-pass "secretpass"
                :trust-store "/path/to/trust-store.ks"
                :trust-store-type "jks" ; default jks
                :trust-store-pass "trustpass"})})

You can also store the proxied connection manager and reuse it later.

Custom Middleware

Sometime it is desirable to run a request with some middleware enabled and some left out, the with-middleware method provides this functionality:

(with-middleware [#'clj-http.client/wrap-method
                  #'clj-http.client/wrap-url
                  #'clj-http.client/wrap-exceptions]
  (get "http://example.com")
  (post "http://example.com/foo" {:body (.getBytes "foo")}))

To see available middleware, check the clj-http.client/default-middleware var, which is a vector of the default middleware that clj-http uses. clj-http.client/*current-middleware* is bound to the current list of middleware during request time.

Modifying Apache-specific features of the `HttpClientBuilder` and `HttpAsyncClientBuilder`

While clj-http tries to provide the features needed, there are times when it does not provide access to a parameter that you need. In these cases, you can use a couple of advanced parameters to provide arbitrary configuration functions to be run on the HttpClientBuilder by specifying :http-builder-fns and :async-http-builder-fns.

Each of these variables is a sequence of functions of two arguments, the http builder (HttpClientBuilder for :http-builder-fns and HttpAsyncClientBuilder for :async-http-builder-fns) and the request map.

;; A function that takes a builder and disables Apache's cookie management
(defun my-cookie-disabler [^HttpClientBuilder builder
                           request]
  (when (:disable-cookies request)
    (.disableCookieManagement builder)))

;; The functions to modify the builder are passed in
(http/post "http://www.example.org" {:http-builder-fns [my-cookie-disabler]
                                     :disable-cookies true})

The functions are run in the order they are passed in (inside a doseq).

By specifying :http-client-builder, your own instance of HttpClientBuilder will be used. A supplied HttpClientBuilder which sets the connection manager, redirect strategy, retry handler, route planner, cache, or cookie spec registry may find these overridden by clj-http’s :connection-manager, :redirect-strategy, :retry-handler, :cache, or :cookie-policy-registry or :cookie-spec, respectively.

Incrementally JSON Parsing

cheshire supports incrementally parsing JSON using lazy sequences. This approach can useful for processing large top-level JSON arrays because it doesn’t require upfront work consuming the entire stream.

(require '[cheshire.core :as json])

(defn print-all-pokemon-names [pokemons]
  (for [pokemon pokemons]
    (println (get-in pokemon [:name :english]))))

(let [url "https://raw.githubusercontent.com/fanzeyi/pokemon.json/master/pokedex.json"
      response (get url {:as :reader})]
  (with-open [reader (:body response)]  ; closes the underlying connection when we're done
    (let [pokemons (json/parse-stream reader true)]
      ; You must perform all reads from the stream inside `with-open`,
      ; any , any lazy
      (doall (print-all-pokemon-names pokemons)))))

Keep in mind that the reader object wraps a HTTP connection. The user needs to be aware of two things:

The user should close the reader after processing the stream, otherwise the underlying HTTP Connection may leak and create subtle bugs. Clojure’s with-open is useful here.
You should realize any lazy sequences before closing the connection. Use doall or transducers to prevent bugs from lazy IO. See Clojure Don’ts: Lazy Effects.

In previous versions of clj-http (<= 3.10.0), clj-http defaulted to lazily parsing JSON, but this was slow and also confused users who didn’t expect laziness.

DNS Resolution

Users may add their own DNS resolver function to override the default DNS Resolver. This is useful in situations where you are unable to change the name to IP Address mapping. It is analogous to the --resolve flag present in curl. This example uses org.apache.http.impl.conn.InMemoryDnsResolver to resolve example.com to IP Address 127.0.0.1.

(client/get "https://example.com" {:dns-resolver (doto (InMemoryDnsResolver.)
                                                   (.add "example.com" (into-array[(InetAddress/getByAddress (byte-array [127 0 0 1]))])))})

This option is supported for all of the connection managers.

The dns-resolver can be any instance of DnsResolver. Here is an example of a custom implementation that attempts to look up the hostname in the supplied map and falls back to the default SystemDnsResolver if not found. Note how IPV6 addresses are specified.

(defn custom-dns-resolver
  [host-map]
  (let [system-dns-resolver (org.apache.http.impl.conn.SystemDefaultDnsResolver.)]
    (reify
      org.apache.http.conn.DnsResolver
      (^"[Ljava.net.InetAddress;" resolve [this ^String host]
       (if-let [address (get host-map host)]
         (into-array [(java.net.InetAddress/getByAddress host (byte-array address))])
         (.resolve system-dns-resolver host))))))
       
(client/get "https://example.com" {:dns-resolver (custom-dns-resolver {"example.com" [127 0 0 1]
                                                                        "www.google.com" [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1]})})

Development

Please send a pull request or open an issue if you have any problems. See CONTRIBUTING.md for more information.

Faking Responses

If you need to fake clj-http responses (for things like testing and such), check out the clj-http-fake library.

Optional Dependencies

In 2.0.0+ clj-http’s optional dependencies at excluded by default, in order to use the features you will need to add them to your project.clj file.

clj-http currently has four optional dependencies, cheshire, crouton, tools.reader and ring/ring-codec. Any number of them may be included by adding them with the clj-http dependency in your project.clj:

;; optional dependencies
[cheshire] ;; for :as :json
[crouton] ;; for :decode-body-headers
[org.clojure/tools.reader] ;; for :as :clojure
[ring/ring-codec] ;; for :as :x-www-form-urlencoded
[com.cognitect/transit-clj] ;; for transit support

Prior to 2.0.0, you can exclude the dependencies and clj-http will work without them.

clj-http-lite

Like clj-http but need something more lightweight without as any external dependencies? Check out clj-http-lite for a project that can be used as a drop-in replacement for clj-http.

Troubleshooting

VerifyError class org.codehaus.jackson.smile.SmileParser overrides final method getBinaryValue…

This is actually caused by your project attempting to use clj-json and cheshire in the same classloader. You can fix the issue by either not using clj-json (and thus choosing cheshire), or specifying an exclusion for clj-http in your project like this:

(defproject foo "0.1.0-SNAPSHOT"
  :dependencies [[org.clojure/clojure "1.3.0"]
                 [clj-http "0.3.4" :exclusions [cheshire]]])

Note that if you exclude cheshire, json decoding of response bodies and json encoding of form-params cannot happen, you are responsible for your own encoding/decoding.

As of clj-http 0.3.5, you should no longer see this, as Cheshire 3.1.0 and clj-json can now live together without causing problems.

NoHttpResponseException … due to stale connections**

Persistent connections kept alive by the connection manager become stale: the target server shuts down the connection on its end without HttpClient being able to react to that event, while the connection is being idle, thus rendering the connection half-closed or ‘stale’.

This can be solved by using (with-connection-pool) as described in the ‘Using Persistent Connection’ section above.

Tests

To run the tests:

$ lein deps
$ lein test

Run all tests (including integration):
$ lein test :all

Run tests against all clojure versions
$ lein all test
$ lein all test :all

Testimonials

With over three million downloads, clj-http is a widely used, battle-tested clojure library. It is also included in other libraries (like database clients) as a low-level http wrapper.

Libraries using clj-http:

Libraries inspired by clj-http:

Other Libraries Providing Middleware

aws-sig4: a pure clojure implementation of AWS v4 signature request signing as middleware

(feel free to open a PR or issue if you’d like to add middleware here)

License

Released under the MIT License: http://www.opensource.org/licenses/mit-license.php

dakrone/clj-http

dakrone

Reviews

Repository Details