For my uses it's great that it has both test suite mode and individual invocation mode. I use it to execute a test suite of HTTP requests against a service in CI.
I'm not a super big fan of the configuration language, the blocks are not intuitive and I found some lacking in the documentation assertions that are supported.
Overall the tool has been great, and has been extremely valuable.
I started using interface testing when working on POCs. I found this helps with LLM-assisted development. Tests are written to directly exercise the HTTP methods, it allows for fluidity and evolution of the implementations as the project is evolving.
I also found the separation of testing very helpful, and it further enforces the separation between interface and implementation. Before hurl, the tests I wrote would be written in the test framework of the language the service is written in. The hurl-based tests really help to enforce the "client" perspective. There is no backdoor data access or anything, just strict separation betwen interface, tests and implementation :)
Maintainer here, thanks for the feedbacks. 6-7 years ago, when we started working on Hurl, we started with a JSON then a YAML file format. We gradually convinced ourself to write a new file format and I completely understand that it might feel weird. We tried (maybe not succeeded!) to have something simple for the simple case...
I'm really interested by issues with the documentation: it can always be improved and any issues is welcome!
Hurl is awesome. A while back I ported a small web service from Python to Rust. Having rigorous tests of the public API is amazing; a language-independent integration test! I was able to swap it out with no changes to the public API or website.
Worth mentioning that using Hurl in Rust specifically gives you a nice bonus feature: integration with cargo test tooling. Since Hurl is written in Rust, you can hook into hurl-the-library and reuse your .hurl files directly in your test suite. Demo: https://github.com/perrygeo/axum-hurl-test
Yeah love Hurl, we stared using it back in 2023-09.
We had a test suite using Runscope, I hated that changes weren't versioned controlled. Took a little grunt work and I converted them in Hurl (where were you AI?) and got rid of Runscope.
Now we can see who made what change when and why. It's great.
I think the idea is nice, but I am struggling for why I should use it. I write using Django, which has plenty of hooks for testing within the framework. Why switch to a tool which is blind to my backend and is going to create more work to keep in sync? At minimum, I lose the ability to easily drop into my debugger to inspect why a result went wrong.
There is probably something to be said for keeping a hard boundary between the backend and testing code, but this would require more effort to create and maintain. I would still need to run the native test suite, so reaching out to an external tool feels a little weird. Unless it was just to ensure an API was fully generic enough for people to run their own clients against it.
> Why switch to a tool which is blind to my backend and is going to create more work to keep in sync? At minimum, I lose the ability to easily drop into my debugger to inspect why a result went wrong.
I don't use hurl but I've used other tools to write language agnostic API tests (and I'm currently working on a new one) so here's what I like about these kinds of tests:
- they're blind to the implementation, and that's actually a pro in my opinion. It makes sure you don't rely on internals, you just get the input and the output
- they can easily serve as documentation because they're language agnostic and relatively easy to share. They're great for sharing between teams in addition to or instead of an OpenAPI spec
- they actually test a contract, and can be reused in case of a migration. I've worked on a huge migration of a public API from Perl to Go and we wanted to keep relatively the same contracts (since the API was public). So we wrote tests for the existing Perl API as a non-regression harness, and could keep the exact same tests for the Go API since they were independent from the language. Keeping the same tests gave us greater confidence than if we had to rewrite tests and it was easy to add more during the double-run/A-B test period
- as a developer, writing these forces you to switch context and become a consumer of the API you just wrote, I've found it easier to write good quality tests with this method
It's just an alternative to Postman and similar so you don't have to start a whole damn electron window just to test a few http requests. It's somewhere between a curl script and Postman, so it hits the right spot for many.
We used Hurl to go from a ktor web server to a spring boot rewrite (Java/Kotlin stack). It was a breeze to have a kind of specifications test suite independent of the server stack and helped us a lot in the transition.
Another benefit is we built a Docker image for production and wanted to have something light and not tight to the implementation for integration tests.
I must say, the sample section[1] does an excellent job of making a case for the tool, especially to people who are inclined to make a snap judgement about the usefulness of the tool within the first 5 minutes (I'm sometimes guilty of this).
So, myself and many folks I know have taken to writing tests in the form of ".http" files that can be executed by IDE extensions in VS Code/IDEA.
Those basically go in the form
POST http://localhost:8080/api/foo
Content-Type: application/json
{ "some": "body" }
And then we have a 1-to-1 mapping of "expected.json" outputs for integration tests.
We use a bespoke bash script to run these .http file with cURL, and then compare the outputs with jq, log success/failure to console, and write "actual.json"
Can I use HURL in a similar way? Essentially an IDE-runnable example HTTP request that references a JSON file as the expected output?
And then run HURL over a directory of these files?
You can use hurl in this way. I have projects with a test directory of hurl files, one hurl file per test case. The cases can run one or more http requests. The hurl file can reference external files, capture values from responses for subsequent requests, validate status and outputs. Hurl has various test runner modes and will optionally output overall test results in various parsable formats if you have a larger reporting framework that you would like to hook into.
Obviously better IDEs integration, support for gRPC, Websocket would be very cool.
A favorite of mine is to be available through official `apt`: there has been some work but it's kind of stuck. The Debian integration is the more difficult integration we have to deal. It's not Debian fault, there are a lot of documentation but we've struggled a lot and fail to understand the process.
I took a lot of inspiration from this project when designing my own HTTP testing tool[0]. We needed to be able to run hundreds of tests quickly, and in parallel. If that is something you need and you like Hurl, then you might like Nap also.
The main difference is that nap works off of YAML. You can get an idea of how it works by clicking on "The Basics"[0] on the left-hand side of the page.
This looks interesting. Longtime user of the Vscode-restclient, but have been moving over to httpyac lately for the scripting and cli use. Will take a look to see if hurl is a good fit.
One annoying thing I've found in testing these tools is that a standard hasn't emerged for using the results of one request as input for another in the syntax of `.http` files. These three tools for instance have three different ways of doing it:
* hurl uses `[Captures]`
* Vscode-restclient does it by referencing request names in a variable declaration (like: `@token = {{loginAPI.response.body.token}}`).
* While httpyac uses `@ref` syntax.
From a quick round of testing it seems like using the syntax for one might break the other tools.
Guilty to have created yet-another-format for HTTP client! To "mitigate" this issue, you can use `hurlfmt` (distributed along `hurl`) that would allow you to export a Hurl file to JSON. You could then go from this JSON to another... It's not magic but it can help if you're going to change from Hurl to another thing.
No worries, it's also interesting to see different peoples approaches to the best syntax for this. Exporters/importers do make life a bit easier I suppose.
I don't know what the mechanism/incentive for getting a standard would be either. Probably most likely would be if there was one clear "winner" that everyone else felt the need mirror.
In any case, appreciate the reply and the tool. Good luck with it.
It’s not nearly as lightweight, and one of the major dealbreakers for postman and equivalents (even ignoring all the drama with postman) is that you have to import and export the data in the client in order to get some text file you can just commit to repo. For my team, that’s a dealbreaker, because it means that people write entire suites of stuff, and never commit them, meaning other people end up doing the same work over and over.
Ahhh, ok. I think I’m thinking about Insomnia, which is basically (in fact?) a fork of Postman. Anyways, that fact is what made Postman a dealbreaker for me, even before the drama. Another thing I like about Rest Client is that the configuration is just a text file, so bearer token etc can be updated via script that runs in a loop.
Rest Client has a few cons though, like request chaining.
I was using Rest Client and was very happy with it, but once I needed Rest Client to use my computer's NO_PROXY env variable to avoid using the proxy for a certain url, and I found it was not possible to do that with Rest Client. That's the only reason I had to look for an alternative tool. After an analysis, I liked Bruno and Hurl. I didn't try hurl yet.
https://github.com/mistweaverco/kulala.nvim is an another restish (it can do gRPC to) plugin for neovim. It is intended to be compatible with a Jetbrains as much as possible.
(After I have seen the IntelliJ one from a colleague I was searching for one like that in neovim. That's the best one I found. It's not perfect, but it works.
Edit: The tool from OP looks very neat though. I will try it out. Might be a handy thing for a few prepared tests that I run frequently
yep, I've played with Hurl and find it nice but recently have been leaning into the .http stuff more. IntelliJ has it built in, there's the plugin you linked, and then for CLI i've used httpYac. No "vendor lock in", really easy to share with copy & paste or source control.
With nice editor integration (especially emacs), hurl is a good postman replacement.
Kinda niche, but I wrapped libhurl to make it really easy to make an AWS Lambda availability monitor out of a hurl file https://gitlab.com/manithree/hurl_lambda
You should probably be looking at the Cargo.toml file(s) (for direct dependencies at least) instead of the lock file as the lock file will include dependencies used for dev/testing.
Hurl has been great for testing in my RAD templating web server project. Like dm03514 says itt, 'The hurl-based tests really help to enforce the "client" perspective.' It's packaged for 3 application environments including a docker image (x2 archs, x3 oses) and with Hurl its easy to ensure the tests pass at the client level in all three environments.
It would be nice to have fancy-regex; today I tried to write a regex to match a case like this ~ <link href="/assets/reset.css\\?hash=(.*)" integrity="\\1" rel="stylesheet"> ~ but the regex crate (and thus hurl asserts) can't do backreferences so I guess I'll just live without checking that these two substrings match.
I wish there was some way to test streamed updates / SSE. Basically, open a connection and wait, then run some other http requests, then assert the accumulated stream from the original connection. https://github.com/Orange-OpenSource/hurl/discussions/2636
The one thing I never understood about the Hurl format is why the response status code assertion happens at the request section and not under the `[Asserts]` section. I wonder what the rationale behind that is.
Snapshots diff current with previous output and I only have to accept or reject the diff. I don't have to write the expected response myself. Snapshots can also stub out parts of the response that are not determistic.
> Snapshots can also stub out parts of the response that are not determistic.
TIL! The way I knew to do it was to have a mock implementation that behaved like the real thing, expect for data/time/uuids/..., where there was just a placeholder.
Snapshot tests being able to "mask" those non-deterministic parts sounds cool!
tbh, that seems pretty close to what I would call snapshot testing already. What people usually do with it is using it for more broadly compared to API testing (for example, I currently use it to test snapshots of a TUI application I am developing) - i.e. you can use it whenever your test is of the form "I have something that I can print in some way and that should look the same until I explicitely want it to look differently in the future". There are a bit more bells and wizzles - For example, it is nice that one does not have to write the initial snapshots oneself. You write the test, execute, it creates the resulting file, then you review and commit that - handy little workflow!
Is that POST in the readme sending the password in the query params? Is this shorthand or literally adding them to the params?
I don't really feel the need for a curl replacement. In the past I've used httpie which is pretty slick but I end up falling back to writing tests in python using requests library.
Maybe I'm not the target audience here, but I should still say something nice I guess. It's nice that it's written in Rust, and open source tooling is in need of fresh projects ever since everyone started bunkering up against the AI monolith scraping all their work. We should celebrate this kind of project, I just wish I had a use for it.
The POST in the README is going to send the params in the request body "url form encoded" like a form in a web page. There are more samples on the doc site [1].
Regarding curl, Hurl is just adding some syntax to pass data from request to request and add assert to responses. For a one time send & forget request, curl is the way, but if you've a kind of workflow (like accessing an authentified resource) Hurl is worth a try. Hurl uses libcurl under the hood and you've an option `--curl` to get a list of curl commands.
> The POST in the README is going to send the params in the request body "url form encoded" like a form in a web page.
Is there a different POST request in the readme or are you saying that this example is going to send the "user" and "password" params in the request body?
That seems really surprising to me - how would you then send a POST request that includes query string parameters? The documentation on form parameters [1] suggests there's an explicit syntax for sending form-encoded request parameters
I see it more as a Postman replacement than curl. When I’m working on a set of APIs, I can quickly write a Hurl file with different combinations that I’m working on. There are usually editor integrations to run individual requests. Then I can share the same Hurl file to my team or commit it in the repo.
This is interesting. I’m wondering how programmable this is. Would this project (or any similar ones) be able to POST a json payload with a field set to “now()”?
If I want to open a modal & check something, could it simulate clicking on open modal button? Or is it first level for now (without any support for interactions / javascript magic)
Hurl works only on the HTTP layer, there is no JavaScript engine. If your modal open a form that will trigger some kind of XHR, you'll "simulate" the HTTP traquets that the form modal could have done. Hurl is not a kind of Playwright for instance.
Can you share "fixtures" amongst tests? For example, I wouldn't want to have to copy/paste the signup and login process for each type of user across hundreds of tests.
What about test isolation? Are people using something else to "prime" the service before/after running these tests?
Is there any program code that is not based on plain text? Punch cards maybe? For the value of "plain text" that includes a programming language with its own syntax and grammar, like Hurl.
For my uses it's great that it has both test suite mode and individual invocation mode. I use it to execute a test suite of HTTP requests against a service in CI.
I'm not a super big fan of the configuration language, the blocks are not intuitive and I found some lacking in the documentation assertions that are supported.
Overall the tool has been great, and has been extremely valuable.
I started using interface testing when working on POCs. I found this helps with LLM-assisted development. Tests are written to directly exercise the HTTP methods, it allows for fluidity and evolution of the implementations as the project is evolving.
I also found the separation of testing very helpful, and it further enforces the separation between interface and implementation. Before hurl, the tests I wrote would be written in the test framework of the language the service is written in. The hurl-based tests really help to enforce the "client" perspective. There is no backdoor data access or anything, just strict separation betwen interface, tests and implementation :)
I'm really interested by issues with the documentation: it can always be improved and any issues is welcome!
Worth mentioning that using Hurl in Rust specifically gives you a nice bonus feature: integration with cargo test tooling. Since Hurl is written in Rust, you can hook into hurl-the-library and reuse your .hurl files directly in your test suite. Demo: https://github.com/perrygeo/axum-hurl-test
We had a test suite using Runscope, I hated that changes weren't versioned controlled. Took a little grunt work and I converted them in Hurl (where were you AI?) and got rid of Runscope.
Now we can see who made what change when and why. It's great.
Loved Runscope it served it's purpose until something came along that that offered the same + version control.
There is probably something to be said for keeping a hard boundary between the backend and testing code, but this would require more effort to create and maintain. I would still need to run the native test suite, so reaching out to an external tool feels a little weird. Unless it was just to ensure an API was fully generic enough for people to run their own clients against it.
I don't use hurl but I've used other tools to write language agnostic API tests (and I'm currently working on a new one) so here's what I like about these kinds of tests:
- they're blind to the implementation, and that's actually a pro in my opinion. It makes sure you don't rely on internals, you just get the input and the output
- they can easily serve as documentation because they're language agnostic and relatively easy to share. They're great for sharing between teams in addition to or instead of an OpenAPI spec
- they actually test a contract, and can be reused in case of a migration. I've worked on a huge migration of a public API from Perl to Go and we wanted to keep relatively the same contracts (since the API was public). So we wrote tests for the existing Perl API as a non-regression harness, and could keep the exact same tests for the Go API since they were independent from the language. Keeping the same tests gave us greater confidence than if we had to rewrite tests and it was easy to add more during the double-run/A-B test period
- as a developer, writing these forces you to switch context and become a consumer of the API you just wrote, I've found it easier to write good quality tests with this method
Another benefit is we built a Docker image for production and wanted to have something light and not tight to the implementation for integration tests.
[1] https://github.com/Orange-OpenSource/hurl?tab=readme-ov-file...
Those basically go in the form
And then we have a 1-to-1 mapping of "expected.json" outputs for integration tests.We use a bespoke bash script to run these .http file with cURL, and then compare the outputs with jq, log success/failure to console, and write "actual.json"
Can I use HURL in a similar way? Essentially an IDE-runnable example HTTP request that references a JSON file as the expected output?
And then run HURL over a directory of these files?
If that's possible, I guess the only thing I'd request is interopability with the REST Client ".http" files that VS Code/JetBrains IDE's support then.
UPDATE: Found it, looks like you can do it via the below
So that just leaves the IDE integration bit.Is your expected.json the actual response body, or is it an object containing body, status, header values, and time-taken, etc?
I really like it because it serves 3 purposes:
- API docs/examples that you can interact with
- Test cases
- Manually invoking API endpoints when working on the underlying code, in an iterative loop
Where do you see hurl in the next 2 years?
A favorite of mine is to be available through official `apt`: there has been some work but it's kind of stuck. The Debian integration is the more difficult integration we have to deal. It's not Debian fault, there are a lot of documentation but we've struggled a lot and fail to understand the process.
[1]: https://github.com/Orange-OpenSource/hurl/issues/366
If I find time I could throw a spec file + ci/cd workflow to get you going that way too
[0] https://naprun.dev
[0] https://naprun.dev/the-basics/
One annoying thing I've found in testing these tools is that a standard hasn't emerged for using the results of one request as input for another in the syntax of `.http` files. These three tools for instance have three different ways of doing it:
* hurl uses `[Captures]`
* Vscode-restclient does it by referencing request names in a variable declaration (like: `@token = {{loginAPI.response.body.token}}`).
* While httpyac uses `@ref` syntax.
From a quick round of testing it seems like using the syntax for one might break the other tools.
[1]: https://hurl.dev/docs/capturing-response.html
[2]: https://github.com/Huachao/vscode-restclient
[3]: https://httpyac.github.io/guide/metaData.html#ref-and-forcer...
I don't know what the mechanism/incentive for getting a standard would be either. Probably most likely would be if there was one clear "winner" that everyone else felt the need mirror.
In any case, appreciate the reply and the tool. Good luck with it.
Conway's Law in action, ladies and gentlemen.
https://marketplace.visualstudio.com/items?itemName=humao.re...
Which is a banger VS Code extension for all sorts of http xyz testing.
It is targeted toward more postman crowd though. May not be as lightweight.
Rest Client has a few cons though, like request chaining.
I was using Rest Client and was very happy with it, but once I needed Rest Client to use my computer's NO_PROXY env variable to avoid using the proxy for a certain url, and I found it was not possible to do that with Rest Client. That's the only reason I had to look for an alternative tool. After an analysis, I liked Bruno and Hurl. I didn't try hurl yet.
(After I have seen the IntelliJ one from a colleague I was searching for one like that in neovim. That's the best one I found. It's not perfect, but it works.
Edit: The tool from OP looks very neat though. I will try it out. Might be a handy thing for a few prepared tests that I run frequently
It gives you full control of constructing requests and assertions because test scenarios may include arbitrary JavaScript.
Kinda niche, but I wrapped libhurl to make it really easy to make an AWS Lambda availability monitor out of a hurl file https://gitlab.com/manithree/hurl_lambda
https://github.com/Orange-OpenSource/hurl/blob/master/Cargo....
https://blog.jetbrains.com/idea/2022/12/http-client-cli-run-...
It would be nice to have fancy-regex; today I tried to write a regex to match a case like this ~ <link href="/assets/reset.css\\?hash=(.*)" integrity="\\1" rel="stylesheet"> ~ but the regex crate (and thus hurl asserts) can't do backreferences so I guess I'll just live without checking that these two substrings match.
I wish there was some way to test streamed updates / SSE. Basically, open a connection and wait, then run some other http requests, then assert the accumulated stream from the original connection. https://github.com/Orange-OpenSource/hurl/discussions/2636
TIL! The way I knew to do it was to have a mock implementation that behaved like the real thing, expect for data/time/uuids/..., where there was just a placeholder. Snapshot tests being able to "mask" those non-deterministic parts sounds cool!
We used it very often a couple of years ago. Will try hurl.
The deficiencies in huel with client state management is not easy to fix.
What I'd like is full client state control with better variable management and use.
For my last project I used Python to write the tests, which appears to work well initially. Dunno how well it will hold up for ongoing maintenance.
Sounds a lot like Emacs' restclient-mode, and I can absolutely see the appeal for those which don't already have an Emacs session open.
You make GET request to server with any of supported crawlers and obtain result in JSON
https://github.com/rumca-js/crawler-buddy/
Supports request, selenium, Httpx, curl cffi, etc
https://github.com/zackify/legible
I don't really feel the need for a curl replacement. In the past I've used httpie which is pretty slick but I end up falling back to writing tests in python using requests library.
Maybe I'm not the target audience here, but I should still say something nice I guess. It's nice that it's written in Rust, and open source tooling is in need of fresh projects ever since everyone started bunkering up against the AI monolith scraping all their work. We should celebrate this kind of project, I just wish I had a use for it.
Regarding curl, Hurl is just adding some syntax to pass data from request to request and add assert to responses. For a one time send & forget request, curl is the way, but if you've a kind of workflow (like accessing an authentified resource) Hurl is worth a try. Hurl uses libcurl under the hood and you've an option `--curl` to get a list of curl commands.
[1]: https://hurl.dev/docs/samples.html
Is there a different POST request in the readme or are you saying that this example is going to send the "user" and "password" params in the request body?
> POST https://example.org/login?user=toto&password=1234
That seems really surprising to me - how would you then send a POST request that includes query string parameters? The documentation on form parameters [1] suggests there's an explicit syntax for sending form-encoded request parameters
[1]: https://hurl.dev/docs/request.html#form-parameters
[1]: https://github.com/Orange-OpenSource/hurl/issues/4151
“Get data from the last log entry in <file> and post it to <url>”
[1]: https://hurl.dev/docs/manual.html#netrc
première fois que je vois qqch de cool sortir d'orange.
I don’t think the DSL is significantly easier than a PL and it’s more limited to?
Is it because of raw speed or ease of reading the DSL?
What about test isolation? Are people using something else to "prime" the service before/after running these tests?