Show HN: I built a tool to get instant test results

by _cfl0 • 4 years ago

96 63 4 years ago

I got sick of the old software development loop: Change code -> Run tests -> wait -> wait some more -> look at failures.

I decided to build a tool that will enable you to: Change code -> look at failures.

No wait time, no explicit test running.

Under the hood:

  - Runs the whole test suite and collects code coverage per test.

  - For each auto file save, analyzes the changes on the tests.

  - Runs changed tests in the background.

  - Display results, the loop time from change to test results is approx 250ms.

Instead of: code -> alt+tab -> arrow up -> rerun all the tests -> wait ... -> test results

code -> alt+tab -> test results!

Check it out! <https://github.com/nabaz-io/nabaz>

newbieuser• 4 years ago

most test runners probably only support running tests that change. it's a good idea to collect this in one tool. good job

_cfl021• 4 years ago

Appreciated! Would love to hear you’re feedback after you ran it also :)

jwhiles• 4 years ago

I'm a bit confused how this differs from the existing watch modes that are built into test runners I'm familiar with, such as Jest. Is this something that I'm used to in TypeScript, but which just doesn't exist in most language's ecosystems?

_cfl021• 4 years ago

Well, I would say the biggest difference is while the watch mode you described runs all the tests for every change.

hypertest only runs the tests the were impacted by the changed code. Resulting is blazing fast performance.

I’d say it’s even fast than apples alt+tab animation most of the time.

volkk2• 4 years ago

hypertest only runs the tests the were impacted by the changed code

with all due respect, this is also a solved problem with test watchers. unless you mean this is smart enough to not just run unit tests that are directly related to the file changed, but also understands code changes from the perspective of downstream dependencies?

Cthulhu_2• 4 years ago

Well, I would say the biggest difference is while the watch mode you described runs all the tests for every change.

This is not true though. Might depend on the test tooling used, but most of the ones I've used only run changed tests (either directly or through change detection via git, e.g. https://nx.dev/concepts/affected)

asdev• 4 years ago

I'm confused too. You can add regex patterns for both filenames and test names in Jest to run only the tests you want in watch mode. This tool just makes it so you don't have to input that configuration? Doesn't seem like a huge value add.

sgeisenh• 4 years ago

For folks using bazel, there is also [bazel-watcher][1]. Also worth checking out [watchman][2] for this style of workflow in general.

[1] https://github.com/bazelbuild/bazel-watcher [2]: https://facebook.github.io/watchman/

_cfl021• 4 years ago

Bazel is awesome, it just runs way too many tests to be fast enough

hamandcheese• 4 years ago

It sounds like your bazel packages might be too large then.

Bazel (and bazel watcher) really shines when you have lots of small packages.

isatty• 4 years ago

--test_filter?

daveguy• 4 years ago

I like the concept and it seems that "run them all and see what changed" is a great model to avoid missing any side effects from a code change. However, if you're running all the tests on every change, doesn't that mean the responsiveness is necessarily greater than 250ms if your test suite is longer than 250ms?

Regardless, running tests in the background to eliminate human delay is good. It would be nice if the (~test time)*(test change likelihood) of the tests were used to order tests based on fastest relevant feedback.

A code->test change likelihood map may also be used to reveal previously unexpected dependencies. Also ~test time could be difficult (see halting problem and windows time remaining estimates).

This is good work.

jasamer2• 4 years ago

This tool doesn't run all the tests, only the ones affected by the code changes since the last run. It figures out which tests to run by initially running all tests, and storing the code coverage of each test.

It doesn't need to rely on "test change likelihood" - if a code change is outside of the code coverage of the test, it doesn't affect the test.

dboreham• 4 years ago

So...like gradle?

marginalia_nu• 4 years ago

If a code change is outside of the code coverage of the test, it doesn't affect the test.

This assertion questionable, the extent it is true it is language specific.

Here is a counterexample in Java:

  // foo.java 
  public class Foo {
    public static final int LENGTH = 10;
  }

  // bar.java 
  public class Bar {
    public int getFooLength() {
      return Foo.LENGTH;
    }
  }

Code coverage tools will not highlight Foo as being touched when Bar.getFooLength is invoked, even though a test asserting that getFooLength() is 10 will fail or succeed depending on Foo.LENGTH. The reason for this is that Foo.LENGTH is inlined in Bar. If you for example wanted to patch a new version of the value, you need to supply Bar.class to affect the change; patching Foo alone would do nothing.

C and C++ likewise do not satisfy this requirement because macros.

_cfl021• 4 years ago

The ordering feature is actually being built as we speak. analyzing liklihood can still be beneficial to cut down time even more.

If I would get a break from hn comments I could finish it today lol ;)

whycombinetor• 4 years ago

1. Does this mean I have to disable IDE autosave so it doesn't attempt to run tests when the project is in an intermediate, invalid state?

2. How does it detect what tests need to be re-run when a given file changes?

_cfl021• 4 years ago

1. No, if build fails hypertest will just discard the result. autosave is even prefered.

2. The first run and every subsequent one collects code coverage which is saved in an sqlite db. Every file change is compared to the code coverage.

Although this approach isnt perfect, The idea here is being fast. I always rerun the entire suite just before committing the code.

mzs• 4 years ago

That's very clever, thanks for this!

jasamer2• 4 years ago

I don't think #1 matters, because it runs again once it's in a valid state again. Your test will fail while in an invalid state, but that's to be expected.

As the author explained in another comment, it initially generates the code coverage for all tests. When code is changed, only the tests that cover that code are rerun.

_cfl021• 4 years ago

Furthermore, it will still display the build error + the previous failure list.

So you can always rely on that background terminal for actual test failure reasons.

gjadi2• 4 years ago

Slow is smooth, smooth is fast.

Though I understand the intent, I think it has the risk to run into a run-hack-run-hack programming style.

I'm currently reading Code Complete which advises against this practice. Instead, one should strive to write code that works from the start instead of hoping it to work eventually.

_cfl021• 4 years ago

Love that mantra! Dont really agree, it’s not intended for that.

I use it as I would run tests through a terminal normally. Built this because I hate running tests.

gjadi2• 4 years ago

Got it.

Very much like the idea to check the code coverage to find out which tests you should run. First time I saw it. Is it common in other tools?

0xbadcafebee• 4 years ago

This is a great idea! Even better: write a Playwright/Codegen[1] test and have that execute continuously (or put a button in the IDE to trigger it). That way you can iterate on multiple code files but see how it impacts one end-to-end test in real time! (I'm developing a feature right now where I'm doing that manually and it would be nice to automate)

[1] https://playwright.dev/docs/codegen-intro

_cfl021• 4 years ago

collab w/ me ;)

hudixt• 4 years ago

We have built something along these lines, Crusher [1] (demo [2]) is an e2e framework, you can trigger test with click of a button. It's built on top of playwright.

We’re thinking of adding watch mode in the coming months - running tests continuously in real-time. Sending you an email to exchange ideas.

P.S.- there might be few typos. It’s work in progress.

[1] https://www.youtube.com/watch?v=Nc-TlgeKBSE [2] https://github.com/crusherdev/crusher

iveqy• 4 years ago

This is great and really close to what I use myself:

while [ 1 ]; do inotifywait .{c,h} t/.{c,h}; make && make check ; done

_cfl021• 4 years ago

Problem being is “make && make check” take way too much time normally.

Aeolun• 4 years ago

This reminds me of Wallaby (except not for Javascript).

klaussilveira• 4 years ago

I have always used:

  watch -n 1 run-test-tool

But I guess this depends on your test tool of choice not being too slow to boot.

_cfl021• 4 years ago

Not only the time to boot, but also if you don’t run tests selectively, you’re already too slow.

talhof8• 4 years ago

Nice! Got an upvote

Liked the name (נב״ז), good luck

_cfl021• 4 years ago

Toda ;)

polyrand• 4 years ago

I do something similar using entr [0][1]. I use it instead of `watch` because I would rather not re-run things if no relevant files have changed.

For example, I run:

  fd -e j2 -e py | entr -c python3 build.py

To rebuild a static site every time the build script[2] (py) or a Jinja2 template (j2) changes.

[0] https://github.com/eradman/entr

[1] https://jvns.ca/blog/2020/06/28/entr/

[2] https://gist.github.com/polyrand/3bed83897658806bd490e1d44df...

nfriedly• 4 years ago

I believe this is similar to how Jest works in watch mode, where it only runs the tests for code that changed.

twic• 4 years ago

This sounds a lot like a couple of old JUnit test runners:

InfiniTest - https://infinitest.github.io/ an dhttps://infinitest.github.io/doc/intellij#how-it-works

JUnit Max - https://web.archive.org/web/20090206151635/http://www.threer... partially surviving as https://junit.org/junit4/javadoc/4.12/org/junit/experimental...

Ruby has something similar:

autotest - https://github.com/grosser/autotest

I think it's a good idea, and i'm surprised it hasn't become a standard feature of IDEs and test runners.

Cthulhu_2• 4 years ago

Jest, probably the most popular JS test runner of today, also has had it since forever.

solardev• 4 years ago

My Jest can run on only changed files, but it still takes forever. Much longer than 250ms

_cfl021• 4 years ago

Had what since forever? Does jest analyze the changes you make and compare them to code coverage? (hint: no)

The key innovation piece here is the speed that the extreme selectively enables you.

grncdr• 4 years ago

Ruby also has crystalball (https://toptal.github.io/crystalball/) which is similar, but not quite the same.

_cfl021• 4 years ago

Main difference is that crystalball still runs all the tests, but runs the recently failed ones first.

Nabaz selects a small subset of affected tests, reducing CPU usage, and more importantly improving speed drastically!

funcDropShadow2• 4 years ago

Just for comparison "sbt ~test" reruns all affected tests whenever some source changes. Affected tests are determined by the dependency analysis of the scala/java compiler.

theptip• 4 years ago

“Test impact analysis” is one name for this general approach. There are a few projects doing this in the Python ecosystem that don’t fully work (Smother, pytest-tia, and some others). Great to see another entry.

Can you cache the coverage DB so this can be run in a CI pipeline?

berkay• 4 years ago

Foresight does this if you're using github actions for CI. https://www.runforesight.com/test-gap-analysis (disclaimer: one of the founders)

_cfl021• 4 years ago

Sure, read docs.nabaz.io (CI solution w/ storage over mongodb)

yardstick2• 4 years ago

hypertest is an open-source CLI tool designed to help you maintain focus on the current coding subtask, built specifically for the easily distracted.

I don’t quite get this proposition. If easily distracted, and wanting to stay focused, wouldn’t you rather want to run tests only on demand and not be interrupted by failing tests after changing just a line or two and still not finished with the rest of your changes.

I can see the benefits of a fast feedback loop. But not sure the pitch of keeping you focused is how i would see it work for me. YMMV.

bagels• 4 years ago

If you are already running tests after every save, and watching for the outcome, but the test takes too long, you might get distracted waiting for the test

yardstick2• 4 years ago

Sure. I would suggest a different workflow rather than trying to optimise the current workflow.

_cfl021• 4 years ago

You aren't interrupted, you choose when to alt+tab to see your test results.

ludovicianul• 4 years ago

For Java devs, Quarkus does this for a while: https://quarkus.io/guides/continuous-testing

Show HN: I built a tool to get instant test results

Related Stories