Snippet generator - don't enforce whitespace rule for numerical parameters

[sebrose]

Summary

#657 has changed the behaviour of the snippet generator such that it no longer behaves as expected with numerical parameters.

Expected Behavior

"Given Sean is 15m from Lucy" -> "Sean is {int}m from Lucy"
"Given a coffee costs $2.50" -> "a coffee costs ${double}" - or maybe {float}, or historically {int}.{int}

Current Behavior

"Given Sean is 15m from Lucy" -> "Sean is 15m from Lucy"
"Given a coffee costs $2.50" -> "a coffee costs $2.{int}"

Possible Solution

Numerical parameters should be bounded by whitespace at one side at least.

• $1.50 -> ${double} or ${int}.{int}
• 15m -> {int}m
• 500ms -> {int}ms
• i18n -> i18n
• $15m -> $15m

Careful consideration of whether punctuation counts as whitespace is needed. "." and "," are used differently by locale.

Context & Motivation

This change results in non-intuitive Cucumber expressions in snippets.

Your Environment

• Version used: Cucumber-JVM 4.7.4
• Operating System and version: Macos Catalina
• Link to your project: https://github.com/cucumber-ltd/cucumber-school-java

[luke-hill]

Good spot. I think the possible solution also makes a lot of sense. And is reasonably easy to interpret (Albeit it might be a bit cumbersome to check the 2 permutations, but don't know yet).

For now a custom one with a letter (Assuming you mean time / distance), would work nicely so \d+\w for example.

[sebrose]

Wouldn't that also match "C3PO"?

\s\d+\w+ would force the integer to be preceded by whitespace.
^\d+\w+ would catch an integer at the start of a line

[luke-hill]

I assumed it was only one letter, so \w would be fine, or you could use a range [] to pass in all valid chars.

It's something I'd happily work on. I know Rien did most of the work alongside aslak, so I might wait to see if they're better placed. But it's quite comprehensively explained by yourself.

[sebrose]

I don't think we should assume a single letter. Consider milliseconds, for example: ms

[aslakhellesoy]

We should complete #771 before tackling this.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs.

[mpkorstanje]

This isn't blocked by #771. The generation of expressions is a different component.

[mpkorstanje]

Numerical parameters should be bounded by whitespace at one side at least.

@sebrose referencing a few style guides I found that units are not always supposed to be attached to the number.

• https://developers.google.com/style/units-of-measure
• https://ukmetric.files.wordpress.com/2017/07/ukma-style-guide.pdf
• https://www.manula.com/manuals/dmp/dmp-writers-style-guide/1/en/topic/units-of-measurement
• https://physics.nist.gov/cuu/Units/checklist.html

Only for angles, percentages and some currencies (USD, but not the Euro) the symbol is attached to the number. So I think all parameters should be bounded on both sides by either whitespace, punctuation, or symbols.

So good examples would be:

"Given Sean is 15 m from Lucy" -> "Sean is {int} m from Lucy"
"Given a coffee costs $2.50" -> "a coffee costs ${double}"

[mpkorstanje]

And writing that in a mono-spaced font makes me think that the style guide might not be applicable here. 😆

[mpkorstanje]

A good counter to only requiring a word boundary on one side would be:

Given I take flight KLM404 --> I take flight KLM404

Rather then:

Given I take flight KLM404 --> I take flight KLM{int}

[mpkorstanje]

I'm going to err on the side of caution for now.

We've got the currency use case covered. Just not the measurements use case.

[marnen]

My take: “KLM404” isn’t an integer; it’s a different thing with its own syntax (just as a phone number is). It should probably not use an {int} parameter type.

OTOH, my experience has been that declaring a custom parameter type is harder than I wish it were (which is why I still write most of my steps with regexes). Perhaps that’s where to focus effort (or perhaps I’m behind the times; I’m not programming full-time these days).

[mpkorstanje]

So I've merged #1108 because its an improvement already.

@marnen perhaps you could check a few of your projects for the [\d.,]+[^\d]+ and [^\d]+[\d.,]+ patterns and see what typical examples might be?

I've checked a few projects myself and with the exception of currency and product codes the pattern simply did not occur. But this is of-course highly domain specific. The only unit of measure I could find was written out in full (e.g. "10 seconds, rather then 10s).

[marnen]

@mpkorstanje I use integers moderately frequently in my step definitions. I’ll find some examples and post links here.