httpcli
This project is a proof of concept of a modern python networking cli which can be simple and easy to maintain using some of the best packages in the python ecosystem:
- click for the foundation of a CLI application. There is also asyncclick that I used in this project which is a tiny wrapper around click to provide asynchronous support.
- rich for pretty printing in the terminal.
- httpx for HTTP protocol stuff.
- anyio for concurrency.
- pytest and pytest-trio for easy testing.
This is not a complete and mature project like httpie but I want to implement some features not present in this beautiful package like:
- HTTP2 support
- more authentication scheme support like digest and oauth2
- easy cookies support
- support of posix signals like SIGINT and SIGTERM
- completion feature
- git "did you mean" like feature
- sse support
Evolution
I'm not quite sure if I will continue improving it without any motivation (sponsoring?) but it is already useful if you want to test it, you just need to have poetry dependency manager and install the project locally (poetry install
). This will install two commands:
http
useful when you don't want the cli to verify server certificate.https
when you need to verify server certificate.
Usage
Hopefully subcommand usage should be straightforward, but I will point some specific cases.
Usage: http [OPTIONS] COMMAND [ARGS]...
HTTP CLI
Options:
--config-file FILENAME A configuration file with options used to
set the cli. Note that the file takes
precedence over the other options.
-t, --timeout FLOAT Time for request to complete, a negative
value means there is no timeout.
--follow-redirects / -N, --no-follow-redirects
flag to decide if http redirections must be
followed
--auth JSON_AUTH A json string representing authentication
information.
--http-version [h1|h2] Version of http used to make the request.
--proxy URL Proxy url.
--version Show the version and exit.
--help Show this message and exit.
Commands:
delete Performs http DELETE request.
download Process download of urls given as arguments.
get Performs http GET request.
head Performs http HEAD request.
install-completion Install completion script for bash, zsh and fish...
options Performs http OPTIONS request.
patch Performs http PATCH request.
post Performs http POST request.
put Performs http PUT request.
sse Reads and print SSE events on a given url.
Global cli configuration
There are some options that can be configured on the root command. These options can be read from a yaml
file using option --config-file
. The config file looks lie the following:
# all options have default values, no need to specify them all
httpcli:
http_version: h2
follow_redirects: true
proxy: https://proxy.com
# timeout may be null to specify that you don't want a timeout
timeout: 5.0
auth:
type: oauth2
flow: password
username: user
password: pass
# for https you also have the verify option to pass a custom certificate used to authenticate the server
verify: /path/to/certificate
Those options can also be configured via environment variables. They are all prefixed with HTTP_CLI_
and they can be in lowercase or uppercase. Here is the same configuration as above but using environment variables:
HTTP_CLI_HTTP_VERSION=h2
HTTP_CLI_FOLLOW_REDIRECTS=true
HTTP_CLI_PROXY=https://proxy.com
HTTP_CLI_TIMEOUT=5.0
# here value is passed as json
HTTP_CLI_AUTH={"type": "oauth2", "flow": "password", "username": "user", "password": "pass"}
HTTP_CLI_VERIFY=/path/to/certificate
Commands
install-completion
This is obviously the first command you will want to use to have subcommand and option autocompletion. You don't need to do that for the two cli http
and https
. Doing it with one will install the other. The current shells supported are bash
, zsh
and fish
. To use autocompletion for subcommands, just enter the first letter and use TAB
key twice. For option autocompletion, enter the first dash and use TAB
twice.
get, head, option, delete
The usage should be pretty straightforward for these commands.
http get --help
Usage: http get [OPTIONS] URL
Performs http GET request.
URL is the target url.
Options:
-c, --cookie COOKIE Cookie passed to the request, can by passed multiple
times.
-H, --header HEADER Header passed to the request, can by passed multiple
times.
-q, --query QUERY Querystring argument passed to the request, can by
passed multiple times.
--help Show this message and exit.
You can play with it using https://pie.dev. Here is a simple example:
http get https://pie.dev/get -c my:cookie -q my:query -H X-MY:HEADER
post, put, patch
There are some subtleties with these commands. I will use post
in the following examples but the same apply to put
and patch
.
json data
If you play with json, in case you only have string values, you can do this:
# here we are sending {"foo": "bar", "hello": "world"} to https://pie.dev/post
http post https://pie.dev/post -j foo:bar -j hello:world
If you need to send other values than strings, you will need to pass the json encoded value with a slightly different syntax, :=
instead of =
.
http post https://pie.dev/post -j number:='2' -j boolean:='true' -j fruits:='["apple", "pineapple"]'
If you have a deeply nested structure you can't write simple in the terminal, you can use of json file instead. Considering we have a file fruits.json with the following content:
[
"apple",
"pineapple"
]
You can use the file like it:
http post https://pie.dev/post -j fruits:@fruits.json
form data
First you need to know that you can't pass form data and json data in the same request. You must choose between the two methods. The basic usage is the following:
https post https://pie.dev/post -f foo:bar -f number:2
If you need to send files, here is what you can do:
# this will send the key "foo" with the value "bar" and the key "photo" with the file photo.jpg
https post https://pie.dev/post -f foo:bar -f photo:@photo.jpg
If you want to send raw data, use the following form:
https post https://pie.dev/post --raw 'raw content'
You can also pass the raw content in a file:
# you can put what you want in your file, just be sure to set the correct content-type
https post https://pie.dev/post --raw @hello.txt
download
You can pass urls as arguments. Files will be downloaded in the current directory. If you wish to change the directory where files should be put, pass the -d
option with the path of the desired destination folder.
# this will downloads two files and put them in the downloads directory of the current user
https download https://pie.dev/image/jpeg https://pie.dev/image/png -d ~/downloads
You can use a file to specify all the resources to download. There should be one url per line. Consider a file urls.txt
having the following content:
https://pie.dev/image/svg
https://pie.def/image/webp
You can download urls from the file and urls from the command line at the same time:
https download https://pie.dev/image/jpeg -f urls.txt
sse
If you want to listen sse events from an endpoint, you can simply do this:
# The sse command will not stop if the data are sent without interruption, which is almost always the case
# with sse, so if you want to stop it, just Ctrl + C ;)
https sse https://endpoint.com/sse
What needs to be improved?
If I were to continue the development of the project, here are the points to review/enhance:
- adapt code to support httpx 1.0 . At the moment of writing it is still in beta, but there is at least one breaking change concerning
allow_redirects
option. - add more authentication schemes, mainly all the oauth2 flows, but may be some others like macaroon...
- support multiple proxy values
- session support
- add CI/CD
- improve code coverage (not 100% yet)
- refactor a bit the code, currently I don't like the structure of my helpers modules. Also auth support can be refactored using this technique I was not aware of when starting this project.
- add autocompletion featurefor other shells like ksh, powershell or powercore
- and probably more... :)