For a more quantitative comparison, the (reduce #(assoc-in ...)) approach
generates O(* (count index-seq) (log_32 (count input-vec))) garbage, while
the (vec (for ...)) approach, with its temporary seq the size of the
vector, and likely wholesale replacement of the vector itself (assuming you
don't keep the old version around after this), generates O(count input-vec)
garbage. They both only traverse the whole vector once, copying in one case
and building up the index-seq in the other, so they're much more even in
speed however large the vector is (both O(count input-vec)).
This is why the assoc-in approach is only really better if the vector is
considerably larger than the number of entries needing updating. Otherwise,
the (vec (for ...)) approach produces simpler, more maintainable code and
should be preferred if you never expect the vectors to grow very large, or
if you expect the proportion of items needing updating to often be a
substantial fraction of "all of them".
On Tue, Jul 30, 2013 at 6:34 PM, Cedric Greevey <cgree...@gmail.com> wrote:
> Note: to repeatedly assoc-in over a seq of indices you'd want
>
> (reduce #(assoc-in %1 [%2 :c] 50) input-vec index-seq)
>
> Replace the key :c and value 50 as needed. If the 50 would actually be
> computed from the index, an expression using %2 would work. If the 50 would
> actually be computed from the existing value associated with the :c (or
> whatever) key, use update-in instead of assoc-in and pass in the function
> that transforms the existing value to the appropriate new value as the last
> argument to update-in.
>
>
>
> On Tue, Jul 30, 2013 at 6:27 PM, Cedric Greevey <cgree...@gmail.com>wrote:
>
>> That's basically the same as the first solution, in terms of how it works
>> under the hood.
>>
>> If the desire is for the output to share structure with the input, not
>> just the maps but the vector as well, then you need to do something a bit
>> more complicated. First you need to find the map you want, and return its
>> index in the vector (assuming it will be unique if present). Then you need
>> to use assoc-in with the index, like (assoc-in v [idx :c] 50) or whatever.
>>
>> If there may be more than one map that should be adjusted, you could
>> return a seq of indices and repeatedly assoc-in, but I doubt this would
>> give superior efficiency unless the vector was typically huge and the
>> number of maps needing changes typically much smaller. With million-element
>> vectors and two or three maps needing updating it would likely be
>> preferable to the (vec (for ...)) approach. With dozen-element vectors with
>> half a dozen maps needing updating I'd go with (vec (for ...)).
>>
>>
>>
>> On Tue, Jul 30, 2013 at 4:58 PM, Jim - FooBar(); <jimpil1...@gmail.com>wrote:
>>
>>> Hi,
>>>
>>> how about ....
>>>
>>> (vec
>>> (for [{:keys [a b c] :as m} aa]
>>> (if (= a 50) (assoc m :c 3) m)))
>>>
>>> :)
>>> Jim
>>>
>>>
>>>
>>> On 30/07/13 18:53, cej38 wrote:
>>>
>>> Suppose I have a vector of maps (this could also be a vector of records)
>>>
>>> (def aa [{:a 0 :b 0 :c 0} {:a 50 :b 0 :c 0} {:a 100 :b 0 :c 0}])
>>>
>>> and I want to go in an change the value of associated with :c when :a
>>> has a value of 50, so that the final col of maps looks like
>>>
>>> [{:a 0 :b 0 :c 0} {:a 50 :b 0 :c 3} {:a 100 :b 0 :c 0}].
>>>
>>>
>>>
>>> I came up with a function that does this:
>>>
>>> (defn find-assoc-in
>>> [k v mv]
>>> (let [f #(if (= ((first k) %) (first v))
>>> (assoc % (second k) (second v))
>>> %)]
>>> (map f mv)))
>>>
>>>
>>> (find-assoc-in [:a :c] [50 3] aa)
>>>
>>>
>>> The problem with this function is that it potentially does a lot
>>> of unnecessary copying of elements that aren't changed, and thus has a lot
>>> of garbage collection. I would like help coming up with something that
>>> does much less garbage collection.
>>>
>>>
>>> Will someone please point out a better way of doing this?
>>>
>>> --
>>> --
>>> 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/groups/opt_out.
>>>
>>>
>>>
>>>
>>> --
>>> --
>>> 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/groups/opt_out.
>>>
>>>
>>>
>>
>>
>
--
--
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/groups/opt_out.
