如何从curl获取文本格式请求？如何将curl命令转为HTTP请求文本？

Great question! Let's break this down into two clear parts: first, how to pull plain-text HTTP requests directly from curl commands, and second, converting your specific curl example into the hand-written format you’re aiming for.

Curl has built-in tools to show you exactly what it sends to the server—no external tools needed. Here are the most practical methods:

• Verbose Mode (-v):
  Add the -v flag to your curl command, and look for lines starting with > in the output. These lines represent the raw request headers and body being sent. For example:

  curl -v https://some.api.com/$batch -H "curl/7.9.8 (i686-pc-linux-gnu) libcurl 7.9.8 (OpenSSL 0.9.6b) (ipv6 enabled)" -d '{ foo: ["bar" "baz"]}'
  

  The output will include all request details, though you’ll need to ignore the debug lines that don’t start with >.

• Trace Mode for Clean Output:
  For a more focused view, use --trace-ascii - to output a detailed trace, then filter to isolate just the request. A quick one-liner with grep works well:

  curl --trace-ascii /dev/stdout https://some.api.com/$batch -H "..." -d '...' | grep -A 100 '=> Send header'
  

  This will show you the full request headers followed by the body, without extra noise.

First, let’s fix a couple of issues in your original curl command and the target request:

• The -H value you provided is actually a User-Agent string—curl normally sends this automatically, but you’re overriding it here. It’s better to explicitly label it as User-Agent: in the header for clarity.
• The JSON body { foo: ["bar" "baz"]} is invalid (missing a comma between "bar" and "baz").

Corrected Curl Command (for valid JSON): #

curl https://some.api.com/$batch \
-H "User-Agent: curl/7.9.8 (i686-pc-linux-gnu) libcurl 7.9.8 (OpenSSL 0.9.6b) (ipv6 enabled)" \
-H "Content-Type: application/json" \
-d '{"foo": ["bar", "baz"]}'

Corresponding Hand-Written HTTP Request: #

POST /$batch HTTP/1.1
Host: some.api.com
User-Agent: curl/7.9.8 (i686-pc-linux-gnu) libcurl 7.9.8 (OpenSSL 0.9.6b) (ipv6 enabled)
Content-Type: application/json
Accept-Encoding: gzip, deflate, br
Content-Length: 21

{"foo":["bar","baz"]}

If you intended to send form data instead of JSON, the Content-Type would default to application/x-www-form-urlencoded, and the body would be formatted accordingly.

For regular conversions, here’s a reliable process:

1. Identify the HTTP method: POST if using -d/--data; GET by default; use -X for PUT/DELETE/etc.
2. Split the URL: Extract the Host (e.g., some.api.com) and request path (e.g., /\$batch).
3. Map headers: Each -H flag becomes a request header. Remember curl auto-adds headers like Accept-Encoding unless you override them.
4. Validate the body: Fix syntax errors (like missing commas in JSON), calculate Content-Length as the byte count of the body, and set the correct Content-Type.

For semi-automated local conversions, you can use a simple script to clean up curl’s trace output. Here’s a one-liner that extracts just the raw request:

curl --trace-ascii - https://some.api.com/$batch -H "..." -d '...' | awk '/=> Send header/{flag=1;next}/<= Recv header/{flag=0}flag' | sed 's/^ //'

内容的提问来源于stack exchange，提问作者simone