[email protected]

kvc supports Key Value Coding-like queries on common Erlang data structures. A common use case for kvc is to quickly access one or more deep values in decoded JSON, or some other nested data structure. It can also help with some aggregate operations. It solves similar problems that you might want to use a tool like XPath or jQuery for, but it is far simpler and strictly less powerful. It's inspired by Apple's NSKeyValueCoding protocol from Objective-C.

The following common Erlang data structures are supported:

• list()
• dict()
• gb_trees()
• proplist()
• {struct, proplist()} (commonly used in mochijson2)
• {proplist()} (EEP 18)
• map() Erlang 17+

Only the following data types are permitted for keys, and they must be UTF-8 if any type coercion takes place:

• atom()
• binary()
• string()

Another limitation is that it is assumed that the given data structure has a homogeneous key type. For example, if any key is binary(), all keys should be binary().

The following collection operators are supported. Note that the numerics have straightforward implementations and no special care is taken reduce floating point error.

Their native representation is binary. Atom and string are also supported but require an additional type coercion for the look-up.

• <<"@sum">>
• <<"@min">>
• <<"@max">>
• <<"@count">>
• <<"@avg">>
• <<"@distinctUnionOfArrays">>
• <<"@distinctUnionOfObjects">>
• <<"@unionOfArrays">>
• <<"@unionOfObjects">>

Not used in production, but it has a test suite that passes.

If you decide to use this in your production app, you should use lists for paths and try to use the same type as the keys in your data structure.

If you'd like to contribute to kvc, a good implementation for setters is the biggest missing piece.

Two styles of queries are supported, the more performant native interface uses a list of keys for the path. If a string, binary, or atom are given then it will be split on '.' peroids to form this key list.

Simple proplist() example:

%% Native key list of atoms that match the data type (fastest)
wibble =:= kvc:path([foo, bar, baz], [{foo, [{bar, [{baz, wibble}]}]}]).

%% Native key list of binaries, does not match key data type (slower)
wibble =:= kvc:path([<<"foo">>, <<"bar">>, <<"baz">>],
                    [{foo, [{bar, [{baz, wibble}]}]}]).

%% These bare keys must be parsed first (slowest)
wibble =:= kvc:path('foo.bar.baz', [{foo, [{bar, [{baz, wibble}]}]}]).
wibble =:= kvc:path(<<"foo.bar.baz">>, [{foo, [{bar, [{baz, wibble}]}]}]).
wibble =:= kvc:path("foo.bar.baz", [{foo, [{bar, [{baz, wibble}]}]}]).

mochijson2 {struct, proplist()} example:

<<"wibble">> =:= kvc:path([<<"foo">>, <<"bar">>, <<"baz">>],
                     {struct,
                      [{<<"foo">>,
                        {struct,
                         [{<<"bar">>,
                           {struct, [{<<"baz">>, <<"wibble">>}]}}]}}]}).

<<"wibble">> =:= kvc:path('foo.bar.baz',
                     {struct,
                      [{<<"foo">>,
                        {struct,
                         [{<<"bar">>,
                         {struct, [{<<"baz">>, <<"wibble">>}]}}]}}]}).

maps example:

<<"wibble">> = kvc:path("foo.bar.baz",
                   #{<<"foo">> => #{<<"bar">> => #{<<"baz">> => <<"wibble">>}}}).

Collection operator example:

2.0 =:= kvc:path([<<"foo">>,<<"bar">>,<<"baz">>,<<"@avg">>],
                 {struct,
                  [{<<"foo">>,
                   {struct,
                    [{<<"bar">>,
                     {struct, [{<<"baz">>, [1, 2, 3]}]}}]}}]}).

2.0 =:= kvc:path('foo.bar.baz.@avg',
                 {struct,
                  [{<<"foo">>,
                   {struct,
                    [{<<"bar">>,
                     {struct, [{<<"baz">>, [1, 2, 3]}]}}]}}]}).

to_proplist normalization:

[{<<"foo">>, [{<<"bar">>, <<"baz">>}]}] =:=
    kvc:to_proplist({struct,
                     [{<<"foo">>,
                       {struct,
                        [{<<"bar">>, <<"baz">>}]}}]}).