00.README.md

Mango Functions & Values

Inspired by apache/couchdb#3828, a draft proposal for adding functions to Mango that can manipulate the data that goes into the index.

Extends the fields index syntax from:

"fields": [{"fieldname":"sort_dir"}]

to

"fields: [{"key": {
    "$fun_name": { "$arg1":"arg1Value", "$arg2": "arg2Value", …}
}

key is the name of the index key that is present in the result set and that can be used in the selector to query on.

$fun_name is the name of the function. I’m imagining the usual slew of string and array funs, maybe some date handling, etc. Probably modelled after the JS standard lib functions.

$argN and $argNValue are additional arguments passed to the function. This could be references to other fields, or any other JSON properties that make sense for a function.

We can also easily extend this to reproduce the value functionality of M/R views emit(key, value) where Mango today only supports emit(key) effectively, even with the additional functions.

Like so:

map: {
  fields: { ... },
  values: {
    "field_name": {
      "$fun_name": {"$arg1": "$arg1Value",…}
    }
  }
}

Where fieldname is where the function result is stored, and returned in the view result like so:

"docs":[
    {"_id":"doc","_rev":"1-XX","bar":"a b c", "_values": { "field_name": "fun_result" }}
  ],

01.ddoc.json

{
  "language": "query",
  "views": {
    "foo-json-index": {
      "map": {
        "fields": { 
          "bar_exp": {
            "$explode": { "$field":"bar", "$separator": " "}
          }
        },
        "partial_filter_selector": {}
      },
      "reduce": "_count",
      "options": {
        "def": {
          "fields": [
            "bar_exp"
          ]
        }
      }
    }
  }
}

02.doc.json

{"bar": "a b c"}

03.selector.json

{"selector": {"bar_exp": {"$eq": "b"}}, "use_index":"mangofun/foo-json-index"}

04.result.json

{
  "docs":[
    {"_id":"doc","_rev":"1-62a23e7b7f0ded1624f42489e5d0c800","bar":"a b c","bar_exp":"b"}
  ],
"bookmark": "g1AAAAA6eJzLYWBgYMpgSmHgKy5JLCrJTq2MT8lPzkzJBYozA1kgKQ6YVA5QkBEkw5iUlQUAxrsPKA"
}

05.patch.erl.diff

diff --git a/src/mango/src/mango_cursor_view.erl b/src/mango/src/mango_cursor_view.erl
index 68d7c3b62..88ffa8ba3 100644
--- a/src/mango/src/mango_cursor_view.erl
+++ b/src/mango/src/mango_cursor_view.erl
@@ -230,6 +230,8 @@ choose_best_index(_DbName, IndexRanges) ->
     {SelectedIndex, SelectedIndexRanges, _} = hd(lists:sort(Cmp, IndexRanges)),
     {SelectedIndex, SelectedIndexRanges}.
 
+add_virtual_field({Props}, Key, [Value]) ->
+    {lists:append(Props, [{Key, Value}])}.
 
 view_cb({meta, Meta}, Acc) ->
     % Map function starting
@@ -238,9 +240,10 @@ view_cb({meta, Meta}, Acc) ->
     ok = rexi:stream2({meta, Meta}),
     {ok, Acc};
 view_cb({row, Row}, #mrargs{extra = Options} = Acc) ->
+    Key = couch_util:get_value(key, Row),
     ViewRow =  #view_row{
         id = couch_util:get_value(id, Row),
-        key = couch_util:get_value(key, Row),
+        key = Key,
         doc = couch_util:get_value(doc, Row)
     },
     case ViewRow#view_row.doc of
@@ -254,9 +257,12 @@ view_cb({row, Row}, #mrargs{extra = Options} = Acc) ->
             put(mango_docs_examined, get(mango_docs_examined) + 1),
             Selector = couch_util:get_value(selector, Options),
             couch_stats:increment_counter([mango, docs_examined]),
-            case mango_selector:match(Selector, Doc) of
+            {[{VirtualField, _}]} = Selector,
+            Doc1 = add_virtual_field(Doc, VirtualField, Key),
+            case mango_selector:match(Selector, Doc1) of
                 true ->
-                    ok = rexi:stream2(ViewRow),
+                    ViewRow1 =ViewRow#view_row{doc = Doc1},
+                    ok = rexi:stream2(ViewRow1),
                     set_mango_msg_timestamp();
                 false ->
                     maybe_send_mango_ping()
diff --git a/src/mango/src/mango_doc.erl b/src/mango/src/mango_doc.erl
index c22b15544..2038359f9 100644
--- a/src/mango/src/mango_doc.erl
+++ b/src/mango/src/mango_doc.erl
@@ -22,6 +22,7 @@
 
     get_field/2,
     get_field/3,
+    get_field_fun/2,
     rem_field/2,
     set_field/3
 ]).
@@ -408,6 +409,19 @@ get_field(Values, [Name | Rest], Validator) when is_list(Values) ->
 get_field(_, [_|_], _) ->
     bad_path.
 
+get_field_fun(Props, MangoFun) ->
+    {FunName, {Args}} = MangoFun,
+    case FunName of
+      <<"$explode">> -> handle_explode(Props, Args);
+      _ -> bad_path
+    end.
+
+handle_explode({Doc}, Args) ->
+    FieldName = proplists:get_value(<<"$field">>, Args),
+    Separator = proplists:get_value(<<"$separator">>, Args),
+    {_, FieldValue} = lists:keyfind(FieldName, 1, Doc),
+    R = string:split(FieldValue, Separator, all),
+    {fn, R}.
 
 rem_field(Props, Field) when is_binary(Field) ->
     {ok, Path} = mango_util:parse_field(Field),
diff --git a/src/mango/src/mango_native_proc.erl b/src/mango/src/mango_native_proc.erl
index 274ae11de..df0418bc6 100644
--- a/src/mango/src/mango_native_proc.erl
+++ b/src/mango/src/mango_native_proc.erl
@@ -146,13 +146,21 @@ get_index_entries({IdxProps}, Doc) ->
             Values = get_index_values(Fields, Doc),
             case lists:member(not_found, Values) of
                 true -> [];
-                false -> [[Values, null]]
+                false -> case Values of
+                    [{fn, Values1}] -> lists:map(fun(V) -> [[V], null] end, Values1);
+                    _Else -> [[Values, null]]
+                end
             end
     end.
 
-
 get_index_values(Fields, Doc) ->
-    lists:map(fun({Field, _Dir}) ->
+    lists:map(fun({_Field, {[MangoFun]}}) ->
+        case mango_doc:get_field_fun(Doc, MangoFun) of
+            not_found -> not_found;
+            bad_path -> not_found;
+            Value -> Value
+        end;
+        ({Field, _Dir}) ->
         case mango_doc:get_field(Doc, Field) of
             not_found -> not_found;
             bad_path -> not_found;

06.mangofun.sh

#!/bin/sh -x

COUCH=a:a@127.0.0.1:15984

curl -sX DELETE $COUCH/mangofun
curl -sX PUT $COUCH/mangofun?q=1

curl -sX PUT $COUCH/mangofun/_design%2fmangofun --data-binary @ddoc.json
#curl -sX PUT $COUCH/mangofun/_design%2fmangofun2 --data-binary @ddoc2.json
#curl -sX PUT $COUCH/mangofun/_design%2fmangofun3 --data-binary @ddoc3.json

curl -sX PUT $COUCH/mangofun/doc --data-binary @doc.json
#curl -sX PUT $COUCH/mangofun/doc2 --data-binary @doc2.json

curl -sX POST -H content-type:application/json $COUCH/mangofun/_find --data-binary @selector.json
#curl -sX POST -H content-type:application/json $COUCH/mangofun/_find --data-binary @selector2.json
#curl -s $COUCH/mangofun/_design/mangofun3/_view/foo?startkey='"b"'

# curl -sX POST -H content-type:application/json $COUCH/mangofun/_explain --data-binary @selector.json | jq