3 Common Caveats

Here we list a few modules and BIFs to watch out for, and not only from a performance point of view.

3.1 The timer module

Creating timers using erlang:send_after/3 and erlang:start_timer/3 is much more efficient than using the timers provided by the timer module. The timer module uses a separate process to manage the timers, and that process can easily become overloaded if many processes create and cancel timers frequently (especially when using the SMP emulator).

The functions in the timer module that do not manage timers (such as timer:tc/3 or timer:sleep/1), do not call the timer-server process and are therefore harmless.

3.2 list_to_atom/1

Atoms are not garbage-collected. Once an atom is created, it will never be removed. The emulator will terminate if the limit for the number of atoms (1048576 by default) is reached.

Therefore, converting arbitrary input strings to atoms could be dangerous in a system that will run continuously. If only certain well-defined atoms are allowed as input, you can use list_to_existing_atom/1 to guard against a denial-of-service attack. (All atoms that are allowed must have been created earlier, for instance by simply using all of them in a module and loading that module.)

Using list_to_atom/1 to construct an atom that is passed to apply/3 like this

apply(list_to_atom("some_prefix"++Var), foo, Args)

is quite expensive and is not recommended in time-critical code.

3.3 length/1

The time for calculating the length of a list is proportional to the length of the list, as opposed to tuple_size/1, byte_size/1, and bit_size/1, which all execute in constant time.

Normally you don't have to worry about the speed of length/1, because it is efficiently implemented in C. In time critical-code, though, you might want to avoid it if the input list could potentially be very long.

Some uses of length/1 can be replaced by matching. For instance, this code

foo(L) when length(L) >= 3 ->
    ...

can be rewritten to

foo([_,_,_|_]=L) ->
   ...

(One slight difference is that length(L) will fail if the L is an improper list, while the pattern in the second code fragment will accept an improper list.)

3.4 setelement/3

setelement/3 copies the tuple it modifies. Therefore, updating a tuple in a loop using setelement/3 will create a new copy of the tuple every time.

There is one exception to the rule that the tuple is copied. If the compiler clearly can see that destructively updating the tuple would give exactly the same result as if the tuple was copied, the call to setelement/3 will be replaced with a special destructive setelement instruction. In the following code sequence

multiple_setelement(T0) ->
    T1 = setelement(9, T0, bar),
    T2 = setelement(7, T1, foobar),
    setelement(5, T2, new_value).

the first setelement/3 call will copy the tuple and modify the ninth element. The two following setelement/3 calls will modify the tuple in place.

For the optimization to be applied, all of the followings conditions must be true:

The indices must be integer literals, not variables or expressions.
The indices must be given in descending order.
There must be no calls to other function in between the calls to setelement/3.
The tuple returned from one setelement/3 call must only be used in the subsequent call to setelement/3.

If it is not possible to structure the code as in the multiple_setelement/1 example, the best way to modify multiple elements in a large tuple is to convert the tuple to a list, modify the list, and convert the list back to a tuple.

3.5 size/1

size/1 returns the size for both tuples and binary.

Using the new BIFs tuple_size/1 and byte_size/1 introduced in R12B gives the compiler and run-time system more opportunities for optimization. A further advantage is that the new BIFs could help Dialyzer find more bugs in your program.

3.6 split_binary/2

It is usually more efficient to split a binary using matching instead of calling the split_binary/2 function. Furthermore, mixing bit syntax matching and split_binary/2 may prevent some optimizations of bit syntax matching.

        <<Bin1:Num/binary,Bin2/binary>> = Bin,

DO NOT

        {Bin1,Bin2} = split_binary(Bin, Num)

3.7 The '--' operator

Note that the '--' operator has a complexity proportional to the product of the length of its operands, meaning that it will be very slow if both of its operands are long lists:

DO NOT

        HugeList1 -- HugeList2

Instead use the ordsets module:

        HugeSet1 = ordsets:from_list(HugeList1),
        HugeSet2 = ordsets:from_list(HugeList2),
        ordsets:subtract(HugeSet1, HugeSet2)

Obviously, that code will not work if the original order of the list is important. If the order of the list must be preserved, do like this:

        Set = gb_sets:from_list(HugeList2),
        [E || E <- HugeList1, not gb_sets:is_element(E, Set)]

Subtle note 1: This code behaves differently from '--' if the lists contain duplicate elements. (One occurrence of an element in HugeList2 will remove all occurrences in HugeList1.)

Subtle note 2: This code compares lists elements using the '==' operator, while '--' uses the '=:='. If that difference is important, sets can be used instead of gb_sets, but note that sets:from_list/1 is much slower than gb_sets:from_list/1 for long lists.

Using the '--' operator to delete an element from a list is not a performance problem:

        HugeList1 -- [Element]