reactivex repeated skip between - reactive-programming

Question

Suppose I have the following stream of data:
1, 2, 3, a, 5, 6, b, 7, 8, a, 10, 11, b, 12, 13, ...
I want to filter everything between 'a' and 'b' (inclusive) no matter how many times they appear. So the result of the above would be:
1, 2, 3, 7, 8, 12, 13, ...
How can I do this with ReactiveX?

Answer 1

Use scan with initial value b to turn
1, 2, 3, a, 5, 6, b, 7, 8, a, 10, 11, b, 12, 13, ...
into
b, 1, 2, 3, a, a, a, b, 7, 8, a, a, a, b, 12, 13, ...
and then filter out a and b to get
1, 2, 3, 7, 8, 12, 13, ...
In pseudo code
values.scan('b', (s, v) -> if (v == 'a' || v == 'b' || s != 'a') v else s).
filter(v -> v != 'a' && v != 'b');

Answer 2

OK. I'm posting this in case anyone else needs an answer to it. A slightly different setup than I described above just to make it easier to understand.
List<String> values = new List<string>()
{
"1", "2", "3",
"a", "5", "6", "b",
"8", "9", "10", "11",
"a", "13", "14", "b",
"16", "17", "18", "19",
"a", "21", "22", "b",
"24"
};
var aa =
// Create an array of CSV strings split on the terminal sigil value
String.Join(",", values.ToArray())
.Split(new String[] { "b," }, StringSplitOptions.None)
// Create the observable from this array of CSV strings
.ToObservable()
// Now create an Observable from each element, splitting it up again
// It is no longer a CSV string but the original elements up to each terminal value
.Select(s => s.Split(',').ToObservable()
// From each value in each observable take those elements
// up to the initial sigil
.TakeWhile(s1 => !s1.Equals("a")))
// Concat the output of each individual Observable - in order
// SelectMany won't work here since it could interleave the
// output of the different Observables created above.
.Concat();
aa.Subscribe(s => Console.WriteLine(s));
This prints out:
1
2
3
8
9
10
11
16
17
18
19
24
It is a bit more convoluted than I wanted but it works.
Edit 6/3/17:
I actually found a cleaner solution for my case.
List<String> values = new List<string>()
{
"1", "2", "3",
"a", "5", "6", "b",
"8", "9", "10", "11",
"a", "13", "14", "b",
"16", "17", "18", "19",
"a", "21", "22", "b",
"24"
};
string lazyABPattern = #"a.*?b";
Regex abRegex = new Regex(lazyABPattern);
var bb = values.ToObservable()
.Aggregate((s1, s2) => s1 + "," + s2)
.Select(s => abRegex.Replace(s, ""))
.Select(s => s.Split(',').ToObservable())
.Concat();
bb.Subscribe(s => Console.WriteLine(s));
The code is simpler which makes it easier to follow (even though it uses regexes).
The problem here is that it still isn't really a general solution to the problem of removing 'repeated regions' of a data stream. It relies on converting the stream to a single string, operating on the string, then converting it back to some other form. If anyone has any ideas on how to approach this in a general way I would love to hear about it.