Re: Enhance spec/double-in to handle open and half-open intervals?

[Mars0i]

On Thursday, July 21, 2016 at 8:50:21 AM UTC-5, miner wrote:
>
> With a little help from Java, you can make equivalent open intervals for 
> the desired bounds.  For example, 
>
> Also, you can use java.lang.Math/nextUp and nextAfter to get adjacent 
> doubles for your bounds. 
>
> (java.lang.Math/nextUp 1.1) 
> ;=> 1.1003 
>
> (java.lang.Math/nextAfter 1.1 Double/NEGATIVE_INFINITY) 
> ;=> 1.0999 
>
> There are a few tricky situations around the zeroes and infinities so you 
> should read the doc on nextAfter, especially if you’re dealing with extreme 
> doubles. 
>

Cool--very nice to know.  Thanks.  I may use it.

Ideally, I'd rather not have to descend to that level for a spec, because 
the meaning becomes less clear, and I'd rather let built-in language 
constructs hide differences between Clojure and Clojurescript whenever 
possible.  Testing for being > 0.0 and < 1.1 won't involve a meaningful 
difference in different clojure variants, for most purposes.

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Steve Miner]

> (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %)) 
> 
> should be the same as 
> 
>   (s/double-in :min Double/MIN_VALUE :max 1.0) 

I should have mentioned that Double/MIN_VALUE is the smallest positive double 
(just greater than 0.0), not a large negative value.  It’s easy to get confused 
by the fact that Long/MIN_VALUE is extremely negative.

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Mars0i]

On Thursday, July 21, 2016 at 9:51:02 AM UTC-5, Alex Miller wrote:
>
> You can already get open intervals by just omitting :min or :max.
>

I think my terminology may have created some confusion; I probably 
shouldn't have used open/closed.  I meant "open interval" in the sense that 
an open interval is one 
that doesn't include the endpoints.  e.g. all real numbers x such that x > 
0 and x < 1 are the open interval (0, 1).  A half-open interval is one that 
includes one of the endpoints, but not the other one.   

(s/double-in :infinite? false) sort of specifies an open interval in my 
sense, except that there are hard boundaries on what can be represented as 
a double, so I suppose it's really a closed interval.  I don't know.  
That's getting too pedantic even for me. :-)

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Alex Miller]

You can already get open intervals by just omitting :min or :max.

On Thursday, July 21, 2016 at 8:50:21 AM UTC-5, miner wrote:
>
> With a little help from Java, you can make equivalent open intervals for 
> the desired bounds.  For example, 
>
>   (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %)) 
>
> should be the same as 
>
>   (s/double-in :min Double/MIN_VALUE :max 1.0) 
>
> Also, you can use java.lang.Math/nextUp and nextAfter to get adjacent 
> doubles for your bounds. 
>
> (java.lang.Math/nextUp 1.1) 
> ;=> 1.1003 
>
> (java.lang.Math/nextAfter 1.1 Double/NEGATIVE_INFINITY) 
> ;=> 1.0999 
>
> There are a few tricky situations around the zeroes and infinities so you 
> should read the doc on nextAfter, especially if you’re dealing with extreme 
> doubles. 
>
>

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Steve Miner]

With a little help from Java, you can make equivalent open intervals for the 
desired bounds.  For example,

  (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %))

should be the same as

  (s/double-in :min Double/MIN_VALUE :max 1.0)

Also, you can use java.lang.Math/nextUp and nextAfter to get adjacent doubles 
for your bounds.

(java.lang.Math/nextUp 1.1)
;=> 1.1003

(java.lang.Math/nextAfter 1.1 Double/NEGATIVE_INFINITY)
;=> 1.0999

There are a few tricky situations around the zeroes and infinities so you 
should read the doc on nextAfter, especially if you’re dealing with extreme 
doubles.

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Mars0i]

On Wednesday, July 20, 2016 at 10:02:17 PM UTC-5, Alex Miller wrote:
>
> On Wednesday, July 20, 2016 at 9:41:59 PM UTC-5, Mars0i wrote:
>>
>> On Wednesday, July 20, 2016 at 4:32:40 PM UTC-5, Alex Miller wrote:
>>>
>>> You can file a jira if you like, I'm not sure Rich's thoughts on this. 
>>>
>>
>> I understand.  Thanks--will do.
>>  
>>
>>> Also, keep in mind that you can also compose preds and get this with 
>>> slightly more effort now:
>>>
>>> (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %))
>>>
>>
>> Yes, definitely.  Though #(and (> % 0.0) (<= % 1)) seems simpler if one 
>> doesn't really need the NaN and Infinity tests. 
>>
>
> You'll find that the generator for double-in is far better than what 
> you're suggesting, and you should lean on it when doing things slightly 
> differently. 
>
> I didn't try it but I don't think your example would gen at all - you'd 
> need to s/and double? in there too at the beginning and even then it's 
> going to generate random doubles then filter to your range, but most 
> generated values will not be in the range. s/double-in is designed to only 
> generate values in the specified range.
>

Ah.  Thanks.   I'm sure you're right.  I didn't understand the role of the 
spec logic functions generating for testing.  I hadn't thought about the 
generator functionality at all--just validation of real inputs.  I'm still 
feeling my way in the dark with spec.  I needed half-open and closed 
interval tests for user input the very small application that I'm using to 
explore spec, which is how the issue about double-in arose for me.  i.e. I 
have a real use case, but my testing can be pretty simple.

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Alex Miller]

On Wednesday, July 20, 2016 at 9:41:59 PM UTC-5, Mars0i wrote:
>
> On Wednesday, July 20, 2016 at 4:32:40 PM UTC-5, Alex Miller wrote:
>>
>> You can file a jira if you like, I'm not sure Rich's thoughts on this. 
>>
>
> I understand.  Thanks--will do.
>  
>
>> Also, keep in mind that you can also compose preds and get this with 
>> slightly more effort now:
>>
>> (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %))
>>
>
> Yes, definitely.  Though #(and (> % 0.0) (<= % 1)) seems simpler if one 
> doesn't really need the NaN and Infinity tests. 
>

You'll find that the generator for double-in is far better than what you're 
suggesting, and you should lean on it when doing things slightly 
differently. 

I didn't try it but I don't think your example would gen at all - you'd 
need to s/and double? in there too at the beginning and even then it's 
going to generate random doubles then filter to your range, but most 
generated values will not be in the range. s/double-in is designed to only 
generate values in the specified range.
 

>
> Thanks.
>
>
>

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Mars0i]

On Wednesday, July 20, 2016 at 4:32:40 PM UTC-5, Alex Miller wrote:
>
> You can file a jira if you like, I'm not sure Rich's thoughts on this. 
>

I understand.  Thanks--will do.
 

> Also, keep in mind that you can also compose preds and get this with 
> slightly more effort now:
>
> (s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %))
>

Yes, definitely.  Though #(and (> % 0.0) (<= % 1)) seems simpler if one 
doesn't really need the NaN and Infinity tests. 

Thanks.


-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Enhance spec/double-in to handle open and half-open intervals?

[Alex Miller]

You can file a jira if you like, I'm not sure Rich's thoughts on this. 

Also, keep in mind that you can also compose preds and get this with 
slightly more effort now:

(s/and (s/double-in :min 0.0 :max 1.0) #(not= 0.0 %))




On Wednesday, July 20, 2016 at 2:03:28 PM UTC-5, Mars0i wrote:
>
> clojure.spec/double-in defines a spec that tests whether a double is 
> greater than or equal to a minimum value and less than or equal to a 
> maximum value.  This is useful for many purposes, but sometimes you need to 
> test whether a double is greater than a minimum or less than a maximum. 
>  There are many mathematical applications that assume half-open intervals 
> (> min and <= max; >= min and < max) or open intervals (> min and < max). 
>  Example: You are going to divide by the double, so it must be > 0, not 
> just >= 0.  Providing double-in but only allowing it to specify closed 
> intervals seems arbitrary.
>
> It's easy enough to write open-interval and half-open interval tests, 
> either using s/and with s/double-in, or just by defining a function from 
> scratch.  However, if Clojure is going to provide double-in, why make us do 
> that?  Why not have a common interface that adds two optional keywords to 
> specify whether the bounds are open or closed?  This is not a request to 
> add the kitchen sink; it's a natural extension to deal with a common use 
> case.  The alternative would be to define three additional versions of 
> double-in, and people will do that on their own, or define a new more 
> flexible double-in if there's no built-in spec function that does it.
>
>
> I'd suggest a new keyword with one of the following forms: 
>
>
> :min-open, :min-greater, :min-greater-than, :min-strict, :greater-min, 
> :strict-min, etc.,
>
>
> and a corresponding keyword for the max side.
>
>
> Default values should be false, for compatibility with the current 
> definition of double-in.
>
>
> I can add a JIRA ticket if this seems worthwhile.  I wanted to see what 
> people think first.  (I'd have to sign up with the JIRA system, but I 
> assume I'm allowed to do that.)  My apologies if there's already a JIRA 
> ticket for this.  I don't seem to be able to find anything on double-in, 
> but I'm not sure I understand the search syntax.
>

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Enhance spec/double-in to handle open and half-open intervals?

[Mars0i]

clojure.spec/double-in defines a spec that tests whether a double is 
greater than or equal to a minimum value and less than or equal to a 
maximum value.  This is useful for many purposes, but sometimes you need to 
test whether a double is greater than a minimum or less than a maximum. 
 There are many mathematical applications that assume half-open intervals 
(> min and <= max; >= min and < max) or open intervals (> min and < max). 
 Example: You are going to divide by the double, so it must be > 0, not 
just >= 0.  Providing double-in but only allowing it to specify closed 
intervals seems arbitrary.

It's easy enough to write open-interval and half-open interval tests, 
either using s/and with s/double-in, or just by defining a function from 
scratch.  However, if Clojure is going to provide double-in, why make us do 
that?  Why not have a common interface that adds two optional keywords to 
specify whether the bounds are open or closed?  This is not a request to 
add the kitchen sink; it's a natural extension to deal with a common use 
case.  The alternative would be to define three additional versions of 
double-in, and people will do that on their own, or define a new more 
flexible double-in if there's no built-in spec function that does it.


I'd suggest a new keyword with one of the following forms: 


:min-open, :min-greater, :min-greater-than, :min-strict, :greater-min, 
:strict-min, etc.,


and a corresponding keyword for the max side.


Default values should be false, for compatibility with the current 
definition of double-in.


I can add a JIRA ticket if this seems worthwhile.  I wanted to see what 
people think first.  (I'd have to sign up with the JIRA system, but I 
assume I'm allowed to do that.)  My apologies if there's already a JIRA 
ticket for this.  I don't seem to be able to find anything on double-in, 
but I'm not sure I understand the search syntax.

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.