Auto-generated descriptions

Testing RSpec
1

Hey all,

I’m looking for ways to optimize rspec and came upon something
interesting. If I remove handling for auto-generated descriptions (the
thing that allows you to say “specify { 5.should == 5 }” with no
description string) we get an average 35% performance increase per
expectation. 35%. That’s huge.

So - how bad do you think this would suck to remove that feature? Are
you using it yourself?

All thoughts welcome.

Thanks,
David

2

On Fri, Jul 4, 2008 at 1:45 PM, David C. [email protected]
wrote:

All thoughts welcome.
To be clear, this is for taking a spec like:

it { 2.should == 2 }

and then when running with the specdoc formatter, getting

  • should == 2

I personally don’t use specdoc format very often anyway. My intuition
is that writing specs without a description (which I frequently use)
is optimized for programmer clarity, so auto-generating a description
for it is less important. I understand that the description would be
used in a number of other places, including error messages, but in the
case of error messages I just use line numbers.

Basically, auto-generating descriptions are valuable, but fall way
short of withholding a 35% performance improvement.

Pat

3

On Fri, Jul 4, 2008 at 4:45 PM, David C. [email protected]
wrote:

So - how bad do you think this would suck to remove that feature? Are you
using it yourself?

Personally, I love this feature, and would not like to see it go.
One of my heuristics for a good spec is that it either a) has a clear
spec doc string; or b) auto-formats to a clearly understandable doc
string. I run spec -fs periodically to verify this. The fact that I
can I can write a sizable subset of my specs in a very DRY way thanks
to the auto-formatting feature is one of my favorite little perks of
using RSpec.

–
Avdi

Home: http://avdi.org
Developer Blog: Avdi Grimm, Code Cleric
Twitter: http://twitter.com/avdi
Journal: http://avdi.livejournal.com

4

On Fri, Jul 4, 2008 at 4:45 PM, David C. [email protected]
wrote:

So - how bad do you think this would suck to remove that feature? Are you
using it yourself?

I’m not, but would it be impractical to extract it out into some sort
of module or helper and tell people to include it in their config.*
block if they want to use it?

5

On Jul 4, 2008, at 4:45 PM, David C. wrote:

So - how bad do you think this would suck to remove that feature?
Are you using it yourself?

All thoughts welcome.

Wasn’t even aware that feature existed!

6

David C. wrote:

All thoughts welcome.

Thanks,
David

rspec-users mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/rspec-users

David,
I personally like that feature, but I honestly don’t find myself using
it too much. 35% is a lot.

One suggestion would be to keep the functionality but move it to a
different keyword so that the common “it”/“specify” case is made
faster. That way we could have our cake and eat it too. I’m not too
sure on what keyword could be used though. “Expect” comes to mind but I
think that is because of Jay Fields expectations framework. Which, BTW,
if we added support for that to coexist within rspec example groups then
the need for rspec to support the feature natively would diminish. So
that is another option that may be worth investigating.

Could you maybe post the diff of your changes to turn off that feature?
I’d be interested in seeing how much speed improvements I get with my
slower spec suites.

Thanks,
Ben

7

On Jul 4, 2008, at 10:54 PM, Mikel L. wrote:

+1

config.enable_auto_generating_descriptions = true

35% is 35%…

Mikel

I would think autoloading rspec (with Module#autoload) would be a big
speed up for features that aren’t used…

Besides - where are the benchmarks?

Scott

8

+1. This is a smart.

9

On Sat, Jul 5, 2008 at 12:18 PM, Steve E. [email protected] wrote:

On Fri, Jul 4, 2008 at 4:45 PM, David C. [email protected] wrote:

So - how bad do you think this would suck to remove that feature? Are you
using it yourself?
I’m not, but would it be impractical to extract it out into some sort
of module or helper and tell people to include it in their config.*
block if they want to use it?

+1

config.enable_auto_generating_descriptions = true

35% is 35%…

Mikel

10

On Jul 04, 2008, at 9:45 pm, David C. wrote:

I’m looking for ways to optimize rspec and came upon something
interesting. If I remove handling for auto-generated descriptions
(the thing that allows you to say “specify { 5.should == 5 }” with
no description string) we get an average 35% performance increase
per expectation. 35%. That’s huge.

So - how bad do you think this would suck to remove that feature?
Are you using it yourself?

I was aware of it but never felt the need to use it

Ashley

–
http://www.patchspace.co.uk/

11

On Fri, Jul 4, 2008 at 10:18 PM, Steve E. [email protected] wrote:

On Fri, Jul 4, 2008 at 4:45 PM, David C. [email protected]
wrote:

So - how bad do you think this would suck to remove that feature? Are you
using it yourself?

I’m not, but would it be impractical to extract it out into some sort
of module or helper and tell people to include it in their config.*
block if they want to use it?

+1

12

On Jul 7, 2008, at 1:47 PM, Ashley M. wrote:

Are you using it yourself?

I was aware of it but never felt the need to use it

Thanks to everyone for the feedback.

I’ve come up w/ a good compromise solution which is to only generate
the description when it’s actually used. Just required moving the call
to generate the description to a different place.

Cheers,
David