Hello,
I am evaluating Connext. It is a big disappointment for my application to find that Connext apparently does not support unbounded sequences. I need to support sequences with bounds in principle up to 2^31 - 1. I can discourage their use, and expect them to be very rare, but cannot just decline to support them. If used, they must cost what they cost. Moreover, they may occur inside of complex types.
1. Am I correct that a User Data Type cannot contain an array with more than 32,767 elements, period?
The only workaround I can see is to do my own serialization and transmit a single large object as a coherent and send-ordered sequence of updates to a smaller object that represents a segment of the serialized byte array. If that is the answer, then I will probably end up not making use of the User Data Type system at all.
2. Is there a better solution?
Thank you,
Chris
On second thought for 2, I'll probably convert the big objects to a relational tables representation whose rows are User Data types in Connext.
Hi Chris,
We are releasing a new RTI Connext DDS version that is compatible wiht unboudned arrays and sequences. It is planned for later this month. I hope this is still useful for you.
Thanks,
Sara
Hi Chris,
As Sara mentioned in Connext DDS 5.1 (and earlier) sequences must have a maximum length. If you specify an unbounded sequence in your IDL then a configurable default is used. Despite this Connext DDS 5.1 does support sending sequences much larger than 32767 elements. You just need to declare the maximum size you want. So the anwer to your first question is no.
For example:
const MAX_LEN = 100000; struct MyTypeWithLargeSeq { sequence<octet, MAXLEN> octet_seq; sequence<long, MAX_LEN> long_seq; };Despite the maximum size, only the actual length will be sent or received. However the memory allocation on the sender and receiver side will allocate each data-samplet to its maximum potential size. This can result on a high use of memory.
With Connext 5.2 we support true unbounded sequence. These sequences have no hard-coded limit. And equally important the memory allocation for each data-sample is trimmed to the actual size. This saves a lot of memory.
Please contact RTI sales if you need access to this feature before the official release of RTI Connext 5.2 later this month. We have evaluation versions of this feature already available both for 5.1 and 5.2 but you need to be explicitely enabled to access these versions.
Gerardo
Hello Chris,
My name is Bert Farabaugh, and I am a Field Applications Engineer for RTI. I am available to help you with any questions you may have related to DDS and our implementation. Please feel free to contact me at :
bertf@rti.com
Thank you,
Bert
Hi Gerardo,
I attended the webast that introduced Connext 5.2 and it is my understanding that unbounded sequence and string are not yet supported for the java language bidning.
Is there any planned release date for this feature to be available also for java binding ?
kind regards
Hi Massimo,
This is considered high priority so it will definitely be in our next release (5.3). We do not have an external date for this yet.
Given the priority we are also considering whether this is something we can include in the 5.2 maintenance release we are plannung for sometime in Q1 2016. I will update this thread once we decide on this.
Gerardo
Hi Gherardo,
thank you for your prompt answer.
Our DDS application is developed part in java and part in c++, so we are forced to define an upper bound value for all the sequences and strings until this feature won't be available also for the java binding.
Kind regards