Question

I would like to apply a function to all the tuples in an ETS table :

The table is a set, each key appears only once.

My table contains only tuples of the same type:

{Key, X, Y, VX, VY}

All values are ::integer()

What i want to do is have a certain value Elapsed and update all
my tuples with a function apply_vector/2

apply_vector({K, X, Y, 0, 0}, _Elapsed) ->
    {K, X, Y, 0, 0};

apply_vector({K, X, Y, VX, VY}, Elapsed) ->
    NewX = X + (VX * Elapsed),
    NewY = Y + (VY * Elapsed),
    {K, NewX, NewY, VX, VY}.

Possible solutions

• If i use ets:foldl, my inserts could be traversed during the fold, and
  result in an infinite (very long) loop.

• I could prepare the new tuples with a ets:foldl, and then insert the whole
  list.

• I could insert to a new table, then replace the old table by the new,
  but i don’t want to limit access to the table with calls to a
  gen_server, the table must be accessible at any time.

• I can not use ets:update_element because I need to read VX and VY values
  to update X and Y.

• I know there are some iterators utilities but no one seems to allow to
  pass a fun.

I need to make this update every 1 – 5 seconds. So, wich solution is
the most efficient with 10 tuples ? With 100 tuples ? With more ?

Thank You !

I Keep a list of boats, the Key is the Boat ID, X and Y are geographic
coordinates, VX and VY represents a vector of movement: a deplacement
for one second. Elapsed is a ratio, a number of seconds since the last
update. The table helps to know at any time the position of each boat.

Answer 1

I probably use ets:foldl for this.

Seems to me like you only need iterative access on this data. So you could just use list of records here.

Another approach (which I would go for) is to put a boat supervisor and create a gen_server for every boat. That way every boat will have its own state and you won’t actually need to traverse any list.

Also checkout qlc http://www.erlang.org/doc/man/qlc.html
qlc let’s you use list comrehensions on ets or mnesia tables. It will probably have the same performance as foldl though.