Hello,
I am a little confused about stream tagging. I read in the documentation
that one tag can be appended to any sample in the data stream so it can
be
something that happens once in a while and it doesn’t have to be at
regular
intervals. That is pretty useful.
Then, when reading about Tagged Stream Blocks it says:
“On the first item of a streamed PDU, there must be a tag with a
specific
key, which stores the length of the PDU as a PMT integer. If anything
else,
or no tag, is on this first item, this will cause the flow graph to
crash!”.
Is it only for the first item or is it PDU after PDU? I mean, a new tag
with the key and length must arrive immediately after one whole PDU is
completed?
If the answer is “yes” I understand it like a vector streaming. What is
the
difference? Then, when is stream tagging useful? I don’t get that from
the
documentation.
What if my custom burst each of them identified with a tag come at
non-defined periods of time? Do I have to develop the block as a
basic/sync
block that read tags of each sample instead and forget about tagged
stream
blocks?
Many thanks in advance. This is a really good community!
Jorge.
PS: At the beginning I found confusing that a Tagged File Sink is not a
Tagged Stream Block given its name but it is only my personal opinion
Hi Jorge,
tagged stream blocks are a specific architecture built atop of the
streaming infrastructure. Not every block that uses tags is a tagged
stream block.
TSB’s are vastly helpful if you’re actually dealin with bursty/packeted
data; the difference to vector processing is that these burst can have
variable length, and that “normal” blocks, which are unaware of the
bursty/PDU nature of the items going through, still work (eg. you can
still multiply a tagged stream block stream with the normal
multiply_const etc).
If you just want to say “hey, this is the sample where XY happened”, no
need to convert your block into a tagged stream block.
Now, regarding your question: yes, when using tagged stream blocks, all
samples must be part of a burst/PDU that started with a tag describing
the PDU’s length. Everything else is illegal.
The motivation behind TSBs was to be able to define blocks that care
about the bursty/packeted nature of items, whilst the rest could
continue to work as it used to.
Best regards,
Marcus
Hey Jorge,
I think Marcus answered your question, but some additional infos:
On 02.03.2015 04:08, Jorge G. wrote:
What if my custom burst each of them identified with a tag come at
non-defined periods of time? Do I have to develop the block as a
basic/sync block that read tags of each sample instead and forget about
tagged stream blocks?
If tags follow the TSB standard practice, all you need to do is derive
your block from tagged_stream_block. It’ll provide you with convenience
functions to handle it.
Many thanks in advance. This is a really good community!
Jorge.
PS: At the beginning I found confusing that a Tagged File Sink is not a
Tagged Stream Block given its name but it is only my personal opinion
The name isn’t perfect, and the ‘tagged stream blocks’ where named such
after the ‘tagged file sink’ existed. Also, you can have streams that
are tagged that are not ‘tagged streams’ in this specific definition.
But this nomenclature just stuck, and we didn’t have anything better at
the time. Names are one of the hardest thing to get right in SW dev,
after all 
M