Uncurl

A Go library to consume a Chrome/Chromium browser "Copy as cURL" string and generate one or more Go *http.Request objects from it

Explanation

In the Chrome or Chromium browser, if you open "Developer tools" and go to the Network tab and navigate somewhere with the browser, you will see a list of requests. Right-clicking one these requests yields a menu with a Copy submenu. One of those options is "Copy as cURL". It allows you to paste a curl command to your terminal or editor, one that reproduces the request if run.

This library accepts that text input and turns it into a Go request. Further Go requests can be generated with different target URLs while maintaining the same header values.

Usage

Initial usage of this package typically follows these steps:

1. Navigate in Chrome/Chromium with dev tools open to network tab, find a request you'd like to extract, right-click and select Copy => Copy as cURL
2. In code, call un, err := uncurl.NewString(str) on the extracted string
3. Generate new requests via uncurl methods, modify request objects, make requests, and extract new URLs from the responses to use in further requests

For a trivial example, let's request the weather in San Diego, and find all the text that looks like a temperature reading in the returned HTTP body.

Example 1: Weather in San Diego

package main

import (
	"bytes"
	"io/ioutil"
	"log"
	"net/http"
	"regexp"

	"github.com/jrefior/uncurl"
)

const tempPattern = `-?\d+(?:\.\d+)? ?°`

const weather = `
curl 'https://www.wunderground.com/weather/us/ca/san-diego' -H 'authority: www.wunderground.com' -H 'upgrade-insecure-requests: 1' -H 'user-agent: Mozilla/5.0 (X11; Fedora; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/79.0.3945.130 Safari/537.36' -H 'accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9' -H 'sec-fetch-site: none' -H 'sec-fetch-mode: navigate' -H 'accept-encoding: gzip, deflate, br' -H 'accept-language: en-US,en;q=0.9' -H 'cookie: usprivacy=foo; s_fid=bar; s_vi=zip' --compressed`

var tempRe *regexp.Regexp

func init() {
	tempRe = regexp.MustCompile(tempPattern)
}

func main() {
	un, err := uncurl.NewString(weather)
	if err != nil {
		log.Fatalf("Failed to initialize uncurl from weather string: %s", err)
	}
	r := un.Request()
	// modify request here if needed
	resp, err := http.DefaultClient.Do(r)
	if err != nil {
		log.Fatalf("Error making request: %s", err)
	}
	if resp.StatusCode != 200 {
		log.Fatalf("Unexpected status code %d", resp.StatusCode)
	}
	b, err := ioutil.ReadAll(resp.Body)
	if err != nil {
		log.Fatalf("Error reading response body: %s", err)
	}
	// often would save the body to a file here for later processing, but in this case let's just
	// search it now
	temps := tempRe.FindAll(b, -1)
	if temps == nil {
		log.Fatal("no temps found")
	}
	log.Printf("Temperatures found: %s\n", bytes.Join(temps, []byte(`, `)))
}

Output:

2020/02/12 07:56:15 Temperatures found: 65°, 48°, 47°, 1.5°, 1.5°, 1.5°, 1.5°, 1.5°, 5°, 41°, 6°, 27°, 81°, 20.9°, 36.4°, 4°, 10°, 5°, 20.9°, 36.4°, 67°, 55°, 67°, 55°, 110°, 80°, 75°, 0°, 30°, 20°, 35°, 58°, 50°, 360°, 360°, 360°, 360°, 360°, 70°, 60°, 36°, 22°, 74°, 70°, 70°, 69°, 3°

Header Value Lists

net/http servers do not split request header values into different elements by comma. This was suspected during review of standard library source code, and then shown via testing. E.g. a curl with this header:

-H 'accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9'

hit a server handler with this code:

func handler(w http.ResponseWriter, r *http.Request) {    
    for k, v := range r.Header {    
        fmt.Printf("%28s %2d\n", k, len(v))    
    }    
}    

which showed the accept header length was 1. A header value slice length of 1 was found for all headers tested, e.g.:

             Accept-Encoding  1
   Upgrade-Insecure-Requests  1
                  User-Agent  1
              Sec-Fetch-Mode  1
             Accept-Language  1
                   Authority  1
                      Accept  1
              Sec-Fetch-Site  1

That behavior is reproduced here on the client side -- no effort is made to split curl -H arguments into different elements.