Hi Chris,

thanks fort he effort.

I agree to merge this into develop if we ensure that we go back to "full 
stable" ASAP.


Am 08.08.19, 16:38 schrieb "Christofer Dutz" <christofer.d...@c-ware.de>:

    Hi all,
    so I just finished refactoring the code generation in the 
"feature/implement-df1-driver" branch. However I think I need to merge the 
things happening there into develop quite soon as now the maven-plugin is no 
longer in sync with the state of develop.
    First of all I cleaned up a little:
    - field --> simple
    - arrayField --> array
    - optionalField --> optional
    Then I added some new fields:
    - checksum (Like an implicit field, but during parsing it reads a value, 
and also calculates the checksum and compares the values and throws an 
exception if they don’t match)
    - padding (If a condition is met, a given input is added ... this is 
generally used in the S7 protocol for word-aligning data)
    - manual (a "simple" field, however do we have full control over 
serialization, deserialization and calculating the length)
    - manualArray (an "array" field, however the developer controls 
serialization, deserialization and calculating the length)
    The "manual" field types can be considered "advanced" features as they do 
require quite a bit of getting used to. But they seem to be doing a nice job.
    I also updated all specs to these changes and added one or two tests to 
test my additions.
    Am 08.08.19, 07:47 schrieb "Christofer Dutz" <christofer.d...@c-ware.de>:
        Hi all,
        so I noticed that new field types require us to re-release the 
build-tools, so I would like to add a few more types:
        - virtual: Generates the getXYZ() method in the POJO
        - manual: Field for which the parse and serialization has to be provided
        - manualArray: like an array version of the manual field
        - fill: which can be used to insert fill bytes in cases like the S7 
where the payloads have to be word-aligned (If an odd number of bytes is 
transferred, a fill byte is added to make it even again) 
        - checksum: which works like an implicit, but also compares the value 
read during reading with the calculated value and throws a Checksum error if 
this doesn't match
        In the mspec format I'll rename the "field" to "simple" as this sort of 
aligns more with the general structure of the rest. Same with the "arrayField" 
which I'd like to call "array".
        Am 07.08.19, 18:23 schrieb "Christofer Dutz" 
            Hi all,
            In the train (Which I didn't miss) I had an idea.
            I'd like to add a "virtual" field which generates a get-method 
without any property to back it in the pojo.
            This can be helpful when working with the model objects.
            It could look something like this:
                [virtual uint 8 'hurz' 'field1.size + field2.size - 5']
            Which generates this in the pojo:
                public short getHurz() {
                        return (short) (getField1().getSize() + 
getField2().getSize() -5);
            For the other case we have a normal field, but want to externally 
or manually map the content form and to the field.
            So how about "externalField" or "manualField" and 
"externalArrayField" or "manualArrayField" for single or multi-value fields.
            For these fields I would suggest to give them two expressions: One 
parsingExpression and one serializationExpression ... the pojo part would be 
identical to the normal field and arrayFields.
            Am 07.08.19, 18:14 schrieb "Christofer Dutz" 
                Hi Julian,
                As all of the elements are "fields" I thought it as a virtual 
field... That's why I suggested that. If you go for "external" it would be an 
external field, which is sort of not quite what it is. And the "whatever name 
it has" - field doesn't have to call external code... It could just be an 
expression which is evaluated... Just in the pojo instead of the io component.
                But I'm not insisting on "virtual" ;-)
                Holen Sie sich Outlook für Android<https://aka.ms/ghei36>
                From: Julian Feinauer <j.feina...@pragmaticminds.de>
                Sent: Wednesday, August 7, 2019 11:49:45 AM
                To: dev@plc4x.apache.org <dev@plc4x.apache.org>
                Subject: Re: [CODE GEN] Extending the mspec to support arrays 
terminated by conditions instead of just "length" and "count", and others ...
                Hi Chris,
                first, thanks for the update here.
                I already checked your results this morning with Volker and 
have to say THANK both of you for your effort, we are pretty close on closing 
this off : )
                For the rest... I agree with your suggestions. Perhaps instead 
of 'virtual' we could use 'external', as we call external code.
                It would be good for us to also generate Interfaces for those 
externals as this would make it easier to add everything in another language, 
if we generate the "skeleton" like that.
                Am 07.08.19, 11:04 schrieb "Christofer Dutz" 
                    Hi all,
                    yesterday Volker and I meet in our codecentric office in 
FFM and worked on a DF1 driver based on mspec and generated code.
                    Here we learned that we need to extend the arrayField to 
support a “terminated” type in addition to the existing “count” (Explicit 
number is specified before the array in the data) or a “length” (the length of 
the payload in bytes is specified before the array in the data).
                    For the DF1 protocol we also need to be able to continue 
adding elements until a termination condition is true (In this case reading the 
0x10 0x03 byte sequence). So I’ll be extending the spec format with this 
feature. I would suggest not to use some sort of termination characters, but to 
call a function which tells the array to read another element or not.
                    Also did we encounter a situation where byte data is 
escaped … so in this case if the data contains the byte 0x10, this has to be 
escaped by duplicating it to 0x10, 0x10. This makes things a little tricky as 
we have to ignore the second 0x10 for the termination condition.
                    Last thing we noticed: for calculating the CRC checksum, it 
would be good to have a new type of field in the spec. One that isn’t used for 
parsing or serializing, but for referencing it (in expressions for example) … 
Not sure how to call it tough … was thinking of “virtual”

Reply via email to