In last week’s Erlang Thursday I concluded with showing how we can take advantage of using ets:select but take advantage of making our queries more expressive.

First we will need a new ETS table, so we start with a new public “Products” table, and do our standard of creating a new process and giving ownership of the table away.

Fun = fun() -> receive after infinity -> ok end end.
% #Fun<erl_eval.20.54118792>
SomeProcess = spawn(Fun).
% <0.52.0>
Products = ets:new(products, [public]).
% 16402
ets:give_away(Products, SomeProcess, []).
% true

Next we will load our “Products” into the table.

In our case, we are just creating a “product” with the “name” as a binary and a “price in CWC” (Common World Currency) as an integer.

[[ ets:insert(Products, {integer_to_binary(X), X}) || X <- lists:seq(1, 100) ]].
% [[true,true,true,true,true,true,true,true,true,true,true,
%   true,true,true,true,true,true,true,true,true,true,true,true,
%   true,true,true,true,true|...]]

As we saw last week, we can manually build up a list of tuples into the match_spec() to run our query, say for items less than 10 CWCs.

ets:select(Products, [{{'$1', '$2'}, [{'<', '$2', 10}], ['$1']}]).
% [<<"8">>,<<"6">>,<<"5">>,<<"3">>,<<"7">>,<<"1">>,<<"4">>,
%  <<"9">>,<<"2">>]

We can also find those item names that are more than 10 CWCs and less than 25 CWCS.

ets:select(Products, [{{'$1', '$2'}, [{'>', '$2', 10}, {'<', '$2', 25}], ['$1']}]).
% [<<"11">>,<<"15">>,<<"23">>,<<"20">>,<<"21">>,<<"14">>,
%  <<"12">>,<<"13">>,<<"16">>,<<"19">>,<<"17">>,<<"18">>,
%  <<"22">>,<<"24">>]

But this isn’t necessarily clear, as we are using numerical values for the items in the tuple, and lists of tuples with lists of tuples inside them.

Enter ets:fun2ms/1 to the rescue.

ets:fun2ms/1 will take a function and will turn that function into a match_spec().

This allows us to write a function that takes a tuple of Product and Cost, and will return the Product if the Cost is less than 10.

ets:fun2ms(fun({Product, Cost}) when Cost < 10 -> Product end).
% [{{'$1','$2'},[{'<','$2',10}],['$1']}]

We can also have compound checks in our guard clauses on the functions we pass to ets:fun2ms/1, such as and clauses,

Between_25_And_35_CWC = ets:fun2ms(fun({Product, Cost}) when Cost > 25, Cost < 35 -> Product end).
% [{{'$1','$2'},[{'>','$2',25},{'<','$2',35}],['$1']}]
ets:select(Products, Between_25_And_35_CWC).
% [<<"30">>,<<"33">>,<<"32">>,<<"29">>,<<"28">>,<<"26">>,
%  <<"34">>,<<"27">>,<<"31">>]

as well as or style clauses.

While this is useful it does have its limitations, as this parse transform on the function, so you can’t use everything that you would be able to with a normal function.

ets:fun2ms(fun({Product, Cost}) when Cost > 90 -> lists:reverse(binary:bin_to_list(Product)) end).
# Error: Unknown error code {122,lists,reverse}
# {error,transform_error}

But then again, the results part of the match_spec(), doesn’t support advanced functionality anyways.

ets:select(Products, [{{'$1', '$2'}, [{'<', 90, '$2'}], [binary:bin_to_list('$1')]}]).
# ** exception error: bad argument
#      in function  binary:bin_to_list/1
#         called as binary:bin_to_list('$1')

But even with its limitations, ets:fun2ms/1 still does a good job to help make our ETS queries more expressive. Not only can we reference a function with expressive variable names over just $X, as well as give guard clauses instead of just guard tuples, but we can also use those variable names in our results as well, and do the basic formatting as part of the function.

And make sure to check back in next week as we continue with looking at the different versions of ets:select.

–Proctor