Smalltalk › Pharo › Pharo Smalltalk Users

Distribution of test method name length

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

6 messages Options

Ben Coman

Distribution of test method name length

Working on the Exercism project to shorten generated test method names,

for comparison I reviewed the length of test method names in Pharo 7.

So just for curiousity value, here is that graph (done in Excel)...

Data generated by...
classes := Object allSubclasses select: [ :cc | cc isKindOf: TestCase class ].
methods := c flatCollect: [ :c | c allMethods ].
tests := methods select: [ :m | m selector beginsWith: 'test' ].
lengths := tests collect: [ :m | m selector size ].
lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow: len; show: ','; show: count ]

cheers -ben

K K Subbu

Re: Distribution of test method name length

Nice graph, Ben! The larger test names (selector size > 100) look more
like sentences than names ;-).

On 21/06/19 6:50 AM, Ben Coman wrote:
>    classes := Object allSubclasses select: [ :cc | cc isKindOf:
> TestCase class ].
>    methods := c flatCollect: [ :c | c allMethods ].

Did you mean classes flatCollect: here?

>    tests := methods select: [ :m | m selector beginsWith: 'test' ].
>    lengths := tests collect: [ :m | m selector size ].

select:thenCollect: can also be used here.

>    lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
> len; show: ','; show: count ]

I find the in: selector very handy for quick commands without having to
use undefined temps. e.g.
````
(Object allSubclasses select: [ :cc | cc isKindOf: TestCase class ]) in:
[ :classes |
(classes flatCollect: [ :c | c allMethods ]) in: [ :methods |
(methods select: [ :m | m selector beginsWith: 'test' ] thenCollect:
[ :m | m selector size ]) in: [:lengths |
lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
len; show: ','; show: count ]]]]
````

Regards .. Subbu

Ben Coman

Re: Distribution of test method name length

On Fri, 21 Jun 2019 at 12:43, K K Subbu <[hidden email]> wrote:

Nice graph, Ben! The larger test names (selector size > 100) look more
like sentences than names ;-).

On 21/06/19 6:50 AM, Ben Coman wrote:
> classes := Object allSubclasses select: [ :cc | cc isKindOf:
> TestCase class ].
> methods := c flatCollect: [ :c | c allMethods ].

Ahh, yes. Blame evolution of the Playground code between when I used it and when I copied it here.

Did you mean "classes" flatCollect: here?

> tests := methods select: [ :m | m selector beginsWith: 'test' ].
> lengths := tests collect: [ :m | m selector size ].

select:thenCollect: can also be used here.

yes. but I was checking each stage as it grew incrementally.

> lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
> len; show: ','; show: count ]

I find the in: selector very handy for quick commands without having to
use undefined temps. e.g.
````
(Object allSubclasses select: [ :cc | cc isKindOf: TestCase class ]) in:
[ :classes |
(classes flatCollect: [ :c | c allMethods ]) in: [ :methods |
(methods select: [ :m | m selector beginsWith: 'test' ] thenCollect:
[ :m | m selector size ]) in: [:lengths |
lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
len; show: ','; show: count ]]]]
````

cool

cheers -ben

Jerry Kott-3

Re: Distribution of test method name length

Interesting…

I am curious about the purpose of this analysis (other than the ‘interesting-ness’ of it). Sure, some names read like sentences, but that beats the ’strcpy()’, doesn’t it? I love that in Smalltalk / Pharo, I don’t have to remember cryptic function names and can make the code optimally verbose (if there is such a thing) to express intent. If that means a method is rather long, so be it.

Jerry Kott

This message has been digitally signed.

PGP Fingerprint:

A9181736DD2F1B6CC7CF9E51AC8514F48C0979A5

On 21-06-2019, at 2:35 AM, Ben Coman <[hidden email]> wrote:

On Fri, 21 Jun 2019 at 12:43, K K Subbu <[hidden email]> wrote:
Nice graph, Ben! The larger test names (selector size > 100) look more
like sentences than names ;-).

On 21/06/19 6:50 AM, Ben Coman wrote:
> classes := Object allSubclasses select: [ :cc | cc isKindOf:
> TestCase class ].
> methods := c flatCollect: [ :c | c allMethods ].

Ahh, yes. Blame evolution of the Playground code between when I used it and when I copied it here.

Did you mean "classes" flatCollect: here?

> tests := methods select: [ :m | m selector beginsWith: 'test' ].
> lengths := tests collect: [ :m | m selector size ].

select:thenCollect: can also be used here.

yes. but I was checking each stage as it grew incrementally.

> lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
> len; show: ','; show: count ]

I find the in: selector very handy for quick commands without having to
use undefined temps. e.g.
````
(Object allSubclasses select: [ :cc | cc isKindOf: TestCase class ]) in:
[ :classes |
(classes flatCollect: [ :c | c allMethods ]) in: [ :methods |
(methods select: [ :m | m selector beginsWith: 'test' ] thenCollect:
[ :m | m selector size ]) in: [:lengths |
lengths asBag keysAndValuesDo: [ :len :count | Transcript crShow:
len; show: ','; show: count ]]]]
````

cool

cheers -ben

Ben Coman

Re: Distribution of test method name length

On Mon, 24 Jun 2019 at 07:10, Jerry Kott <[hidden email]> wrote:

Interesting…

I am curious about the purpose of this analysis (other than the ‘interesting-ness’ of it). Sure, some names read like sentences, but that beats the ’strcpy()’, doesn’t it? I love that in Smalltalk / Pharo, I don’t have to remember cryptic function names and can make the code optimally verbose (if there is such a thing) to express intent. If that means a method is rather long, so be it.

Its an offshoot of working on the Pharo Track of the Exercism project (https://exercism.io/tracks/pharo-smalltalk)

Here we generate test-method names from their canonical data "description" (https://github.com/exercism/problem-specifications/blob/master/exercises/word-count/canonical-data.json)

The original intent of that field was to generate identifiers, but sometimes the language ends up a bit flowery and we ended up with method-name 150 character long. I am in the process of slimming these down.

Someone queried me "what was a recommended identifier length" and I had no clue -- so I thought the Pharo code base would be a good source of data.

Having produced the graph, I thought others might find it mildly interesting. Thats all.

It may be worthwhile reviewing some of the outliers, but that was not its intent.
It was shared just-for-interest. The purpose is certainly not to squeeze Pharo messages down to 6 characters ;)

cheers -ben

P.S. I found the graph more useful showing percentages rather than absolute count.

95% less than 50 characters

99% less than 60 characters

Kasper Osterbye

Re: Distribution of test method name length

Quite interesting! I was wondering about the correlation to number of arguments due to the particularities of the keyword syntax which might give longer method names in the first place.

Best,

Kasper