DimWell
February 26, 2018, 10:08am
1
Hello Experts,
Could you please advise what would be the way to match the bitstring ?
Lets say I have and input and I want to pattern part of it as bitstsing:
@input <<116, 114, 117, 101, 32, 105, 115, 32, 110, 111, 116, 32, 102, 97, 108, 115, 101>>
@true_ <<116, 114, 117, 101>>
@false_ <<102, 97, 108, 115, 101>>
def match_bitsting( <<true_ ::size(32) , rest :: bitstring>>= input) when trues_ == @true_ do
IO.puts "I got: true"
end
def match_bitsting( <<false_ ::size(40) , rest :: bitstring>>= input) when trues_ == @false_ do
IO.puts "I got: false"
end
How should I rewrite @true _ | @false _ or the pattern matching to make it work ?
1 Like
grych
February 26, 2018, 10:16am
2
If I understood the problem properly, you want to match the beginning of the input string? In this case, it is easy:
def match_bitstring("true" <> _rest), do: IO.puts("true")
def match_bitstring("false" <> _rest), do: IO.puts("false")
1 Like
DimWell
February 26, 2018, 10:20am
3
Not really, probably I have to change description a bit…
I have an input that looks like:
<<116, 114, 117, 101, 32, 105, 115, 32, 110, 111, 116, 32, 102, 97, 108, 115, 101, 32, 97, 110, 100, 32, 102, 97, 108, 115, 101, 32, 105, 115, 32, 110, 111, 116, 32, 116, 114, 117, 101>>
I want to do pattern match on:
<<116, 114, 117, 101>> # which is "true" and the rest data loop froward and match according other principles
Furthermore, this just dummy input just to explain the idea, the input can be any text…
idi527
February 26, 2018, 10:23am
4
@input <<116, 114, 117, 101, 32, 105, 115, 32, 110, 111, 116, 32, 102, 97, 108, 115, 101>>
@true_ <<116, 114, 117, 101>>
@false_ <<102, 97, 108, 115, 101>>
def match_bitsting(<<@true_, rest::bits>>) do # bits is an alias for bitstring
IO.puts "I got: true"
match_continue(rest, true)
end
def match_bitsting(<<@false_, rest::bits>>) do
IO.puts "I got: false"
match_continue(rest, false)
end
Would this work for you?
Why are you using bitstring
s anyway? What you have in @input
is a (printable) binary
.
3 Likes
DimWell
February 26, 2018, 10:30am
5
idi527:
Would this work for you?
Yes, this works and so clean code
Somehow I overcomplicated…
DimWell
February 26, 2018, 10:32am
6
Trying to understand how the pattern matching works on bitstring level…
I mean how to use pattern matching in this type of scenario…
I am still in learning and and trying to understanding Elixir / Erlang / Pattern Matching
Thank you very much for help !
NobbZ
February 26, 2018, 11:22am
7
I think what @idi527 meant here, why do you have bitstring
instead of binary
. The first one matches arbitrary length sequences of bits while the latter matches arbitrary length sequences of bytes , and since your provided patterns all seem to stick to a byte boundary the latter might perform better then the first.
DimWell
February 26, 2018, 11:27am
8
Hi Noobz
Hm… When I read some online examples it was mentioned that if you are not sure about provided input then it is better to specify bitstring
rather than binary
…
In my case I assume the input data comes from a some API that comes as stream data.
I thought this is the way to do it.
Am I wrong ?
NobbZ
February 26, 2018, 11:35am
9
If it is streamed data you will probably get your chunks at byte boundary (as this is the level at which your interface transfers the data) and unless the protocol specifies something that is not aligned at byte boundaries, you can safely assume binary
.
BUT If the data is streamed in, you can’t safely use binary pattern matching at all but need to use a different mean of parsing/deserializing the incomming data.
What happens if your current chunk is "tru"
? This does not match "true"
, but might after the next chunk has been received.
3 Likes
DimWell
February 26, 2018, 11:47am
10
Yes, you are right. Thank you for a good point.
DimWell
February 26, 2018, 11:54am
11
I thought to give a try with Genstage & Flow .
Producer:
Something like receives chunks and analyses it according to some logic.
It would split a chunk into invalid and valid parts.
The invalid part would buffered until next chunk & analysis.
The valid part would be available to ConsumerProducer.
ConsumerProducer:
Receives a valid chunk and does the pattern matching.
Does some manipulations and prepares data for the next stage.
Consumer:
Gets final data produces some output.
1 Like
NobbZ
February 26, 2018, 12:13pm
12
This is a valid approach but only if order of processing of the datagrams does not matter.
1 Like
DimWell
February 26, 2018, 12:23pm
13
Probably the implementation itself would be changed later on, but definitely I would need to consider more nuances
At this point in time I just want get better understanding of Genstage, Flow, Stream & How to glue all together
But you already gave me valuable information to think of, thank you very much for that
1 Like
@DimWell Just for the record, what you were trying to do is possible, you just need to use the suffix ‘bits-’ or ‘bytes-’…
<<x::bits-size(16), _::binary>> = "true"
inspect(x) # "tr"
<<x::bytes-size(2), _::binary>> = "true"
inspect(x) # "tr"
;D
2 Likes