Re: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Mars0i 


On Thursday, April 21, 2016 at 2:20:11 PM UTC-5, Mars0i wrote:

> The behavior with strings is simply undocumented anywhere, afaik, though.
>

Sorry, that was wrong even before I added an example to clojure.docs.  The 
conj.io documentation is quite explicit and detailed.

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Alex Miller 
Strings are indexed. They are a special case because String is a Java final 
class that cannot be "extended" to support the core interfaces inside 
Clojure's Java implementation. If Clojure were itself protocol-based (like 
ClojureScript is), the protocol could cover this without the special case.


On Thursday, April 21, 2016 at 10:48:18 AM UTC-5, Mars0i wrote:
>
> The docstring for 'get' says nothing explicit about vectors ("Returns the 
> value mapped to key, not-found or nil if key not present.") but most of us 
> know that vectors can be viewed as associative data structures where 
> indexes are like keys:
>
> (get [\a \b \c] 1) 
> \b
>
> 'get' also works on strings:
>
> (get "abc" 1)
> \b
>
> However, strings are not treated as vectors or maps in other contexts:
>
> (assoc "abc" 1 \Z)
> ClassCastException java.lang.String cannot be cast to 
> clojure.lang.Associative
>
>
Being able to *add* something to an indexed collection is a totally 
different "behavior" than looking something up by index. Strings don't do 
that part.
 

> This combination of behaviors is confusing.  If strings are vectors, then 
> 'assoc' and other similar functions should work with them, and maybe I 
> should be able to do this to index into a string:
>
> ("abc" 1)
> ClassCastException java.lang.String cannot be cast to clojure.lang.IFn
>
> If strings are not vectors, then they seem to be a special case for 
> 'get'.  One might argue that the very terse docstring for 'get' should note 
> its use with vectors, but once you understand the sense in which vectors 
> are like maps, the of 'get' use with vectors is implicitly covered by what 
> the docstring says.  However, the use of 'get' with strings is not 
> implicitly covered by what the docstring says.  If it was, then the last 
> two examples I gave should work.
>
> (I'd be willing to file a ticket on this, but the puzzling behavior is not 
> new, and I suspect that there's something I'm not understanding.)
>

I think the current behavior and docs are correct and I'm not yet convinced 
there is a change needed. (But feel free to keep discussing and maybe I'll 
change my mind. :) 

Alex

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Mars0i 


On Thursday, April 21, 2016 at 11:49:16 AM UTC-5, Andy Fingerhut wrote:
>
> Most doc strings are considered terse by many people.  This appears to be 
> considered a feature by those who maintain Clojure, not a bug, with one 
> reason given by Stuart Halloway in his recent talk on using the scientific 
> method in debugging [1].  Paraphrased from memory it was approximately "if 
> you have a bug in some API you are using, you should read all of the 
> documentation for it.  That is easier if it is short."
>
> If you want longer documentation and/or examples, ClojureDocs.org and 
> conj.io take user submissions for additional documentation, with quite 
> low friction.
>

Yeah, there are tradeoffs.  I understand the value of brief docstrings.  
The current docstring for 'get' is correct for maps and vectors, even 
though the meaning for vectors won't be apparent for novices.  Lots of 
things aren't apparent for novices.  

The behavior with strings is simply undocumented anywhere, afaik, though.  
That seems wrong.  (Correction: I just added an example to clojure.docs.  I 
still think that all of the intended behavior of a function should be at 
least tersely implied by its docstring.)

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread adrian . medina 
That's a good point, but it should be noted that all collections for which 
clojure.core/counted? returns true (they implement clojure.lang.Counted) 
should implement count in constant time. So the design of the Counted 
interface was clearly intended to provide consumers a soft guarantee of 
performance invariability across data structures when required. 

On Thursday, April 21, 2016 at 1:26:26 PM UTC-4, Michael Gardner wrote:
>
> On Apr 21, 2016, at 10:04, James Reeves  > wrote: 
> > 
> > Clojure seems to avoid having functions that have variable performance 
> depending on the data structure they're applied to. 
>
> But not always! (e.g. count)

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread adrian . medina 
That's a good point, but it should be noted that all collections for which 
clojure.core/counted? returns true (they implement clojure.lang.Counted) 
should implement count in constant time. So the design of the Counted 
interface was clearly intended to provide consumers a soft guarantee of 
performance invariability across data structures when necessary.. 

On Thursday, April 21, 2016 at 1:26:26 PM UTC-4, Michael Gardner wrote:
>
> On Apr 21, 2016, at 10:04, James Reeves  > wrote: 
> > 
> > Clojure seems to avoid having functions that have variable performance 
> depending on the data structure they're applied to. 
>
> But not always! (e.g. count)

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Michael Gardner 
On Apr 21, 2016, at 10:04, James Reeves  wrote:
> 
> Clojure seems to avoid having functions that have variable performance 
> depending on the data structure they're applied to.

But not always! (e.g. count)

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread James Reeves 
It doesn't necessarily follow that something that works with
clojure.core/get needs to work with clojure.core/assoc. You can have an
object that implements ILookup but not Associative, for instance.

The problem with using assoc with native JVM data structures like strings
and arrays is that there isn't an efficient way of implementing assoc for
those structures. You have to copy the entire structure, no matter how
large it might be. Clojure seems to avoid having functions that have
variable performance depending on the data structure they're applied to.

For example, trying to "get" on a seq also fails, presumably because seqs
have no efficient way of looking up a value by index. Instead we have
clojure.core/nth, which has predictable O(n) performance.

- James

On 21 April 2016 at 16:48, Mars0i  wrote:

> The docstring for 'get' says nothing explicit about vectors ("Returns the
> value mapped to key, not-found or nil if key not present.") but most of us
> know that vectors can be viewed as associative data structures where
> indexes are like keys:
>
> (get [\a \b \c] 1)
> \b
>
> 'get' also works on strings:
>
> (get "abc" 1)
> \b
>
> However, strings are not treated as vectors or maps in other contexts:
>
> (assoc "abc" 1 \Z)
> ClassCastException java.lang.String cannot be cast to
> clojure.lang.Associative
>
> This combination of behaviors is confusing.  If strings are vectors, then
> 'assoc' and other similar functions should work with them, and maybe I
> should be able to do this to index into a string:
>
> ("abc" 1)
> ClassCastException java.lang.String cannot be cast to clojure.lang.IFn
>
> If strings are not vectors, then they seem to be a special case for
> 'get'.  One might argue that the very terse docstring for 'get' should note
> its use with vectors, but once you understand the sense in which vectors
> are like maps, the of 'get' use with vectors is implicitly covered by what
> the docstring says.  However, the use of 'get' with strings is not
> implicitly covered by what the docstring says.  If it was, then the last
> two examples I gave should work.
>
> (I'd be willing to file a ticket on this, but the puzzling behavior is not
> new, and I suspect that there's something I'm not understanding.)
>
> --
> 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.
>

-- 
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: Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Andy Fingerhut 
This is a special case behavior of clojure.core/get, as you can see in the
RT.java source file at these lines:
https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/RT.java#L750-L755

I have no idea if a JIRA would be accepted for any change on this or not,
but if one was, I suspect it would be an addition to the doc string for
clojure.core/get, mentioning that it also works on strings.

Most doc strings are considered terse by many people.  This appears to be
considered a feature by those who maintain Clojure, not a bug, with one
reason given by Stuart Halloway in his recent talk on using the scientific
method in debugging [1].  Paraphrased from memory it was approximately "if
you have a bug in some API you are using, you should read all of the
documentation for it.  That is easier if it is short."

If you want longer documentation and/or examples, ClojureDocs.org and
conj.io take user submissions for additional documentation, with quite low
friction.

Andy

[1] https://www.youtube.com/watch?v=FihU5JxmnBg

On Thu, Apr 21, 2016 at 8:48 AM, Mars0i  wrote:

> The docstring for 'get' says nothing explicit about vectors ("Returns the
> value mapped to key, not-found or nil if key not present.") but most of us
> know that vectors can be viewed as associative data structures where
> indexes are like keys:
>
> (get [\a \b \c] 1)
> \b
>
> 'get' also works on strings:
>
> (get "abc" 1)
> \b
>
> However, strings are not treated as vectors or maps in other contexts:
>
> (assoc "abc" 1 \Z)
> ClassCastException java.lang.String cannot be cast to
> clojure.lang.Associative
>
> This combination of behaviors is confusing.  If strings are vectors, then
> 'assoc' and other similar functions should work with them, and maybe I
> should be able to do this to index into a string:
>
> ("abc" 1)
> ClassCastException java.lang.String cannot be cast to clojure.lang.IFn
>
> If strings are not vectors, then they seem to be a special case for
> 'get'.  One might argue that the very terse docstring for 'get' should note
> its use with vectors, but once you understand the sense in which vectors
> are like maps, the of 'get' use with vectors is implicitly covered by what
> the docstring says.  However, the use of 'get' with strings is not
> implicitly covered by what the docstring says.  If it was, then the last
> two examples I gave should work.
>
> (I'd be willing to file a ticket on this, but the puzzling behavior is not
> new, and I suspect that there's something I'm not understanding.)
>
> --
> 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.
>

-- 
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.

Are strings vectors? Should 'get' docstring be changed?

2016-04-21 Thread Mars0i 
The docstring for 'get' says nothing explicit about vectors ("Returns the 
value mapped to key, not-found or nil if key not present.") but most of us 
know that vectors can be viewed as associative data structures where 
indexes are like keys:

(get [\a \b \c] 1) 
\b

'get' also works on strings:

(get "abc" 1)
\b

However, strings are not treated as vectors or maps in other contexts:

(assoc "abc" 1 \Z)
ClassCastException java.lang.String cannot be cast to 
clojure.lang.Associative

This combination of behaviors is confusing.  If strings are vectors, then 
'assoc' and other similar functions should work with them, and maybe I 
should be able to do this to index into a string:

("abc" 1)
ClassCastException java.lang.String cannot be cast to clojure.lang.IFn

If strings are not vectors, then they seem to be a special case for 'get'.  
One might argue that the very terse docstring for 'get' should note its use 
with vectors, but once you understand the sense in which vectors are like 
maps, the of 'get' use with vectors is implicitly covered by what the 
docstring says.  However, the use of 'get' with strings is not implicitly 
covered by what the docstring says.  If it was, then the last two examples 
I gave should work.

(I'd be willing to file a ticket on this, but the puzzling behavior is not 
new, and I suspect that there's something I'm not understanding.)

-- 
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.