erts: Implement maps path iterator

4 files changed, 252 insertions, 42 deletions

diff --git a/lib/dialyzer/test/small_SUITE_data/results/maps_sum b/lib/dialyzer/test/small_SUITE_data/results/maps_sum
index bd192bdb93..b29ac77d88 100644
--- a/lib/dialyzer/test/small_SUITE_data/results/maps_sum
+++ b/lib/dialyzer/test/small_SUITE_data/results/maps_sum
@@ -1,4 +1,4 @@
 
-maps_sum.erl:15: Invalid type specification for function maps_sum:wrong1/1. The success typing is (map()) -> any()
+maps_sum.erl:15: Invalid type specification for function maps_sum:wrong1/1. The success typing is (maps:iterator() | map()) -> any()
 maps_sum.erl:26: Function wrong2/1 has no local return
 maps_sum.erl:27: The call lists:foldl(fun((_,_,_) -> any()),0,Data::any()) will never return since it differs in the 1st argument from the success typing arguments: (fun((_,_) -> any()),any(),[any()])
diff --git a/lib/stdlib/doc/src/maps.xml b/lib/stdlib/doc/src/maps.xml
index 8c7270816b..a3afbf8e7f 100644
--- a/lib/stdlib/doc/src/maps.xml
+++ b/lib/stdlib/doc/src/maps.xml
@@ -33,16 +33,32 @@
     <p>This module contains functions for maps processing.</p>
   </description>
 
+  <datatypes>
+    <datatype>
+      <name name="iterator"/>
+      <desc>
+        <p>An iterator representing the key value associations in a map.</p>
+        <p>Created using <seealso marker="#iterator-1"><c>maps:iterator/1</c></seealso>
+        and <seealso marker="#iterator-2"><c>maps:iterator/2</c></seealso>.</p>
+        <p>Consumed by <seealso marker="#next-1"><c>maps:next/1</c></seealso>,
+          <seealso marker="#filter-2"><c>maps:filter/2</c></seealso>,
+          <seealso marker="#fold-3"><c>maps:fold/3</c></seealso> and
+          <seealso marker="#map-2"><c>maps:map/2</c></seealso>.</p>
+      </desc>
+    </datatype>
+  </datatypes>
+
   <funcs>
     <func>
       <name name="filter" arity="2"/>
       <fsummary>Select pairs that satisfy a predicate.</fsummary>
       <desc>
-        <p>Returns a map <c><anno>Map2</anno></c> for which predicate
-          <c><anno>Pred</anno></c> holds true in <c><anno>Map1</anno></c>.</p>
+        <p>Returns a map <c><anno>Map</anno></c> for which predicate
+          <c><anno>Pred</anno></c> holds true in <c><anno>MapOrIter</anno></c>.</p>
         <p>The call fails with a <c>{badmap,Map}</c> exception if
-          <c><anno>Map1</anno></c> is not a map, or with <c>badarg</c> if
-          <c><anno>Pred</anno></c> is not a function of arity 2.</p>
+          <c><anno>MapOrIter</anno></c> is not a map or valid iterator,
+          or with <c>badarg</c> if <c><anno>Pred</anno></c> is not a
+          function of arity 2.</p>
         <p><em>Example:</em></p>
         <code type="none">
 > M = #{a => 2, b => 3, c=> 4, "a" => 1, "b" => 2, "c" => 4},
@@ -76,12 +92,16 @@
       <fsummary></fsummary>
       <desc>
         <p>Calls <c>F(K, V, AccIn)</c> for every <c><anno>K</anno></c> to value
-          <c><anno>V</anno></c> association in <c><anno>Map</anno></c> in
+          <c><anno>V</anno></c> association in <c><anno>MapOrIter</anno></c> in
           any order. Function <c>fun F/3</c> must return a new
           accumulator, which is passed to the next successive call.
           This function returns the final value of the accumulator. The initial
           accumulator value <c><anno>Init</anno></c> is returned if the map is
           empty.</p>
+        <p>The call fails with a <c>{badmap,Map}</c> exception if
+          <c><anno>MapOrIter</anno></c> is not a map or valid iterator,
+          or with <c>badarg</c> if <c><anno>Fun</anno></c> is not a
+          function of arity 3.</p>
         <p><em>Example:</em></p>
         <code type="none">
 > Fun = fun(K,V,AccIn) when is_list(K) -> AccIn + V end,
@@ -169,6 +189,75 @@ false</code>
     </func>
 
     <func>
+      <name name="iterator" arity="1"/>
+      <fsummary>Create a map iterator.</fsummary>
+      <desc>
+        <p>Returns a map iterator <c><anno>Iterator</anno></c> that can
+          be used by <seealso marker="#next-1"><c>maps:next/1</c></seealso>
+          to traverse the keys and values in a map. This call is equivalent
+          to calling <seealso marker="#iterator-2"><c>maps:iterator/2</c></seealso>
+          with an empty map as the options.</p>
+        <p>The call fails with a <c>{badmap,Map}</c> exception if
+          <c><anno>Map</anno></c> is not a map.</p>
+        <p><em>Example:</em></p>
+        <code type="none">
+> M = #{ a => 1, b => 2 }.
+#{a => 1,b => 2}
+> I = maps:iterator(M).
+[{a,1},{b,2}]
+> {K1, V1, I2} = maps:next(I).
+{a,1,[{b,2}]}
+> {K2, V2, I3} = maps:next(I2).
+{b,2,[]}
+> maps:next(I3).
+none</code>
+      </desc>
+    </func>
+
+    <func>
+      <name name="iterator" arity="2"/>
+      <fsummary>Create a map iterator.</fsummary>
+      <desc>
+        <p>Returns a map iterator <c><anno>Iterator</anno></c> that can
+          be used by <seealso marker="#next-1"><c>maps:next/1</c></seealso>
+          to traverse the keys and values in a map. The iterator will behave
+          according to the given <c><anno>Opts</anno></c>.</p>
+        <taglist>
+          <tag>optimize</tag>
+          <item>
+            Configures whether the iterator should be optimized for <c>speed</c>
+            or for <c>memory</c> consumption. Default is <c>speed</c>.
+          </item>
+          <tag>ordered</tag>
+          <item>
+            Configures whether the iterator should return the key value
+            associations ordered on the keys according to erlang term order
+            or not. The default is <c>false</c>.
+          </item>
+        </taglist>
+        <p>Both options can be configured at the same time.</p>
+        <p>The call fails with a <c>{badmap,Map}</c> exception if
+          <c><anno>Map</anno></c> is not a map.</p>
+        <p><em>Example:</em></p>
+        <code type="none">
+> M = maps:from_list([{I,I} || I &lt;- lists:seq(1,50)]).
+#{45 => 45,6 => 6,2 => 2,49 => 49,41 => 41,33 => 33,42 => 42,
+  43 => 43,10 => 10,9 => 9,19 => 19,14 => 14,5 => 5,18 => 18,
+  31 => 31,22 => 22,29 => 29,21 => 21,27 => 27,24 => 24,
+  47 => 47,40 => 40,30 => 30,23 => 23,28 => 28,46 => 46,
+  16 => 16,38 => 38,4 => 4,...}
+> MemoryIter = maps:iterator(M, #{ optimize => memory }), ok.
+ok
+> {K1, V1, _} = maps:next(MemoryIter), {K1, V1}.
+{46,46}
+> OrderedIter = maps:iterator(M, #{ ordered => true, optimize => memory }), ok.
+ok
+> {K2, V2, _} = maps:next(OrderedIter), {K2, V2}.
+{1,1}</code>
+      </desc>
+    </func>
+
+    <func>
       <name name="keys" arity="1"/>
       <fsummary></fsummary>
       <desc>
@@ -188,12 +277,16 @@ false</code>
       <name name="map" arity="2"/>
       <fsummary></fsummary>
       <desc>
-        <p>Produces a new map <c><anno>Map2</anno></c> by calling function
+        <p>Produces a new map <c><anno>Map</anno></c> by calling function
           <c>fun F(K, V1)</c> for every <c><anno>K</anno></c> to value
-          <c><anno>V1</anno></c> association in <c><anno>Map1</anno></c> in
+          <c><anno>V1</anno></c> association in <c><anno>MapOrIter</anno></c> in
           any order. Function <c>fun F/2</c> must return value
           <c><anno>V2</anno></c> to be associated with key <c><anno>K</anno></c>
-          for the new map <c><anno>Map2</anno></c>.</p>
+          for the new map <c><anno>Map</anno></c>.</p>
+        <p>The call fails with a <c>{badmap,Map}</c> exception if
+          <c><anno>MapOrIter</anno></c> is not a map or valid iterator,
+          or with <c>badarg</c> if <c><anno>Fun</anno></c> is not a
+          function of arity 2.</p>
         <p><em>Example:</em></p>
         <code type="none">
 > Fun = fun(K,V1) when is_list(K) -> V1*2 end,
@@ -234,6 +327,35 @@ false</code>
     </func>
 
     <func>
+      <name name="next" arity="1"/>
+      <fsummary>Get the next key and value from an iterator.</fsummary>
+      <desc>
+        <p>Returns the next key-value association in
+          <c><anno>Iterator</anno></c> and a new iterator for the
+          remaining associations in the iterator.
+        </p>
+        <p>
+          If there are no more associations in the iterator,
+          <c>none</c> is returned.
+        </p>
+        <p><em>Example:</em></p>
+        <code type="none">
+> Map = #{a => 1, b => 2, c => 3}.
+#{a => 1,b => 2,c => 3}
+> Iter = maps:iterator(Map).
+[{a,1},{b,2},{c,3}]
+> {_, _, Iter1} = maps:next(Iter).
+{a,1,[{b,2},{c,3}]}
+> {_, _, Iter2} = maps:next(Iter1).
+{b,2,[{c,3}]}
+> {_, _, Iter3} = maps:next(Iter2).
+{c,3,[]}
+> maps:next(Iter3).
+none</code>
+      </desc>
+    </func>
+
+    <func>
       <name name="put" arity="3"/>
       <fsummary></fsummary>
       <desc>
diff --git a/lib/stdlib/src/maps.erl b/lib/stdlib/src/maps.erl
index 0da12cf8e2..2f2dd41979 100644
--- a/lib/stdlib/src/maps.erl
+++ b/lib/stdlib/src/maps.erl
@@ -24,7 +24,7 @@
          map/2, size/1,
          update_with/3, update_with/4,
          without/2, with/2,
-         iterator/1, next/1]).
+         iterator/1, iterator/2, next/1]).
 
 %% BIFs
 -export([get/2, find/2, from_list/1,
@@ -32,10 +32,13 @@
          new/0, put/3, remove/2, take/2,
          to_list/1, update/3, values/1]).
 
--opaque iterator() :: [{term(), term()}] | [[pos_integer() | map()]].
+-opaque iterator() :: {non_neg_integer() | none | list(term()), map()}
+                    | list({term(), term()}).
 
 -export_type([iterator/0]).
 
+-define(IS_ITERATOR(I), (is_list(I) orelse (is_map(element(2,I) andalso (is_list(element(1,I) orelse is_integer(element(1,I)))))))).
+
 %% Shadowed by erl_bif_types: maps:get/2
 -spec get(Key,Map) -> Value when
     Key :: term(),
@@ -44,7 +47,6 @@
 
 get(_,_) -> erlang:nif_error(undef).
 
-
 -spec find(Key,Map) -> {ok, Value} | error when
     Key :: term(),
     Map :: map(),
@@ -197,15 +199,17 @@ get(Key,Map,Default) ->
     erlang:error({badmap,Map},[Key,Map,Default]).
 
 
--spec filter(Pred,Map1) -> Map2 when
+-spec filter(Pred,MapOrIter) -> Map when
       Pred :: fun((Key, Value) -> boolean()),
       Key  :: term(),
       Value :: term(),
-      Map1 :: map(),
-      Map2 :: map().
+      MapOrIter :: map() | iterator(),
+      Map :: map().
 
 filter(Pred,Map) when is_function(Pred,2), is_map(Map) ->
-    maps:from_list(filter_1(Pred, iterator(Map)));
+    maps:from_list([{K,V}||{K,V}<-maps:to_list(Map),Pred(K,V)]);
+filter(Pred,Iterator) when is_function(Pred,2), ?IS_ITERATOR(Iterator) ->
+    maps:from_list(filter_1(Pred, Iterator));
 filter(Pred,Map) ->
     erlang:error(error_type(Map),[Pred,Map]).
 
@@ -222,18 +226,20 @@ filter_1(Pred, Iter) ->
             []
     end.
 
--spec fold(Fun,Init,Map) -> Acc when
+-spec fold(Fun,Init,MapOrIter) -> Acc when
     Fun :: fun((K, V, AccIn) -> AccOut),
     Init :: term(),
     Acc :: term(),
     AccIn :: term(),
     AccOut :: term(),
-    Map :: map(),
+    MapOrIter :: map() | iterator(),
     K :: term(),
     V :: term().
 
 fold(Fun,Init,Map) when is_function(Fun,3), is_map(Map) ->
-    fold_1(Fun, Init, iterator(Map));
+    lists:foldl(fun({K,V},A) -> Fun(K,V,A) end,Init,maps:to_list(Map));
+fold(Fun,Init,Iterator) when is_function(Fun,3), ?IS_ITERATOR(Iterator) ->
+    fold_1(Fun,Init,Iterator);
 fold(Fun,Init,Map) ->
     erlang:error(error_type(Map),[Fun,Init,Map]).
 
@@ -245,16 +251,18 @@ fold_1(Fun, Acc, Iter) ->
             Acc
     end.
 
--spec map(Fun,Map1) -> Map2 when
+-spec map(Fun,MapOrIter) -> Map when
     Fun :: fun((K, V1) -> V2),
-    Map1 :: map(),
-    Map2 :: map(),
+    MapOrIter :: map() | iterator(),
+    Map :: map(),
     K :: term(),
     V1 :: term(),
     V2 :: term().
 
 map(Fun,Map) when is_function(Fun, 2), is_map(Map) ->
-    maps:from_list(map_1(Fun, iterator(Map)));
+    maps:from_list([{K,Fun(K,V)}||{K,V}<-maps:to_list(Map)]);
+map(Fun,Iterator) when is_function(Fun, 2), ?IS_ITERATOR(Iterator) ->
+    maps:from_list(map_1(Fun, Iterator));
 map(Fun,Map) ->
     erlang:error(error_type(Map),[Fun,Map]).
 
@@ -278,28 +286,45 @@ size(Val) ->
       Map :: map(),
       Iterator :: iterator().
 
-iterator(Map) when is_map(Map) ->
-    case maps:size(Map) < 42 of
-        true -> maps:to_list(Map);
-        false -> [[1 | Map]]
+iterator(M) when is_map(M) -> iterator(M, #{});
+iterator(M) -> erlang:error({badmap, M}, [M]).
+
+-spec iterator(Map, Opts) -> Iterator when
+      Map :: map(),
+      Opts :: #{ ordered => boolean(), optimize => speed | memory },
+      Iterator :: iterator().
+iterator(M, Opts) when is_map(M), is_map(Opts) ->
+    case maps:merge(default_iterator_opts(), Opts) of
+        #{ ordered := true, optimize := speed} -> lists:sort(maps:to_list(M));
+        #{ ordered := false, optimize := speed} -> maps:to_list(M);
+        #{ ordered := true, optimize := memory} -> {lists:sort(maps:keys(M)), M};
+        #{ ordered := false, optimize := memory} -> {0, M}
     end;
-iterator(M) ->
-    erlang:error(error_type(M),[M]).
+iterator(M, Opts) when is_map(M) ->
+    erlang:error({badmap, M}, [M, Opts]);
+iterator(M, Opts) ->
+    erlang:error(error_type(M), [M, Opts]).
+
+default_iterator_opts() ->
+    #{ ordered => false, optimize => speed }.
 
--spec next(Iterator) -> {K,V,NextIterator} | 'none' when
+-spec next(Iterator) -> {Key, Value, NextIterator} | 'none' when
       Iterator :: iterator(),
-      K :: term(),
-      V :: term(),
+      Key :: term(),
+      Value :: term(),
       NextIterator :: iterator().
-next([]) -> none;
-next([[Index | Map] | St]) ->
-    case erts_internal:map_get_index(Index, Map) of
-        SubMap when is_map(SubMap) -> next([[1 | SubMap], [Index+1|Map] | St]);
-        [K|V] -> {K,V,[[Index+1 | Map] | St]};
-        none -> next(St)
-    end;
-next([{K, V} | I]) ->
-    {K, V, I}.
+next([{K, V} | T]) ->
+    {K, V, T};
+next([]) ->
+    none;
+next({I, M}) when is_map(M), is_integer(I) orelse is_atom(I) ->
+    erts_internal:map_next(I, M);
+next({[K | T], M}) when is_map(M) ->
+    {K, maps:get(K, M), {T, M}};
+next({[], M}) when is_map(M) ->
+    none;
+next(Iter) ->
+    erlang:error(badarg, [Iter]).
 
 -spec without(Ks,Map1) -> Map2 when
     Ks :: [K],
@@ -332,5 +357,5 @@ with(Ks,M) ->
     erlang:error(error_type(M),[Ks,M]).
 
 
-error_type(M) when is_map(M) -> badarg;
+error_type(M) when is_map(M); ?IS_ITERATOR(M) -> badarg;
 error_type(V) -> {badmap, V}.
diff --git a/lib/stdlib/test/maps_SUITE.erl b/lib/stdlib/test/maps_SUITE.erl
index 42e669a799..0021dd0657 100644
--- a/lib/stdlib/test/maps_SUITE.erl
+++ b/lib/stdlib/test/maps_SUITE.erl
@@ -30,6 +30,7 @@
 -export([t_update_with_3/1, t_update_with_4/1,
          t_get_3/1, t_filter_2/1,
          t_fold_3/1,t_map_2/1,t_size_1/1,
+         t_iterator_1/1, t_iterator_2/1,
          t_with_2/1,t_without_2/1]).
 
 %%-define(badmap(V,F,Args), {'EXIT', {{badmap,V}, [{maps,F,Args,_}|_]}}).
@@ -47,6 +48,7 @@ all() ->
     [t_update_with_3,t_update_with_4,
      t_get_3,t_filter_2,
      t_fold_3,t_map_2,t_size_1,
+     t_iterator_1, t_iterator_2,
      t_with_2,t_without_2].
 
 t_update_with_3(Config) when is_list(Config) ->
@@ -127,6 +129,8 @@ t_filter_2(Config) when is_list(Config) ->
     Pred2 = fun(K,V) -> is_list(K) andalso (V rem 2) =:= 0 end,
     #{a := 2,c := 4} = maps:filter(Pred1,M),
     #{"b" := 2,"c" := 4} = maps:filter(Pred2,M),
+    #{a := 2,c := 4} = maps:filter(Pred1,maps:iterator(M)),
+    #{"b" := 2,"c" := 4} = maps:filter(Pred2,maps:iterator(M)),
     %% error case
     ?badmap(a,filter,[_,a]) = (catch maps:filter(fun(_,_) -> ok end,id(a))),
     ?badarg(filter,[<<>>,#{}]) = (catch maps:filter(id(<<>>),#{})),
@@ -139,6 +143,8 @@ t_fold_3(Config) when is_list(Config) ->
     Tot0 = lists:sum(Vs),
     Tot1 = maps:fold(fun({k,_},V,A) -> A + V end, 0, M0),
     true = Tot0 =:= Tot1,
+    Tot2 = maps:fold(fun({k,_},V,A) -> A + V end, 0, maps:iterator(M0)),
+    true = Tot0 =:= Tot2,
 
     %% error case
     ?badmap(a,fold,[_,0,a]) = (catch maps:fold(fun(_,_,_) -> ok end,0,id(a))),
@@ -151,12 +157,69 @@ t_map_2(Config) when is_list(Config) ->
     #{ {k,1} := 1, {k,200} := 200} = M0,
     M1 = maps:map(fun({k,_},V) -> V + 42 end, M0),
     #{ {k,1} := 43, {k,200} := 242} = M1,
+    M2 = maps:map(fun({k,_},V) -> V + 42 end, maps:iterator(M0)),
+    #{ {k,1} := 43, {k,200} := 242} = M2,
 
     %% error case
     ?badmap(a,map,[_,a]) = (catch maps:map(fun(_,_) -> ok end, id(a))),
     ?badarg(map,[<<>>,#{}]) = (catch maps:map(id(<<>>),#{})),
     ok.
 
+t_iterator_1(Config) when is_list(Config) ->
+
+    %% Small map test
+    M0 = #{ a => 1, b => 2 },
+    I0 = maps:iterator(M0),
+    {K1,V1,I1} = maps:next(I0),
+    {K2,V2,I2} = maps:next(I1),
+    none = maps:next(I2),
+
+    KVList = lists:sort([{K1,V1},{K2,V2}]),
+    KVList = lists:sort(maps:to_list(M0)),
+
+    %% Large map test
+
+    Vs = lists:seq(1,200),
+    M2 = maps:from_list([{{k,I},I}||I<-Vs]),
+    KVList2 = lists:sort(iter_kv(maps:iterator(M2))),
+    KVList2 = lists:sort(maps:to_list(M2)),
+    ok.
+
+t_iterator_2(Config) when is_list(Config) ->
+
+    Vs = lists:seq(1,200),
+    Maps = [#{ a => 1, b => 2 }, maps:from_list([{{k,I},I}||I<-Vs])],
+    Optimize = [speed, memory],
+    Ordered = [true, false],
+
+    [test_iterator(Map, Opt, Ord) || Map <- Maps,
+                                     Opt <- Optimize,
+                                     Ord <- Ordered],
+
+    ok.
+
+test_iterator(Map, Opt, Ord) ->
+    Opts = #{ optimize => Opt, ordered => Ord },
+    test_iterator_1(maps:iterator(Map, Opts), Map, Opts).
+
+test_iterator_1(Iter, OrigMap, Opts) ->
+    erlang:display(Opts),
+    KVs = iter_kv(Iter),
+    case Opts of
+        #{ ordered := true } ->
+            KVs = lists:sort(maps:to_list(OrigMap));
+        #{ ordered := false } ->
+            OrderedKVs = lists:sort(KVs),
+            OrderedKVs = lists:sort(maps:to_list(OrigMap))
+    end.
+
+iter_kv(I) ->
+    case maps:next(I) of
+        none ->
+            [];
+        {K,V,NI} ->
+            [{K,V} | iter_kv(NI)]
+    end.
 
 t_size_1(Config) when is_list(Config) ->
       0 = maps:size(#{}),