Hey Chris,
Sorry for mixing authorship. I didn't check simulator since a while,
hence was not certain if its up to date.
My point is about reusing network layer which PLC4X provides in order to
simulate a device. This is how you normally interact with PLC so it is a
natural way to make an "integration test" at this level. It is not
beyond typical BDD which is more about defining scenarios. How they are
actually executed depend on a BDD toolkits and how they integrate with
test or build environment.
I get that scope of Peter work is to emulate a PLC runtime, that's why I
mentioned IEC grammars (so he can somehow parse or compile entire
program), which is the hardest part.
Once program can be interpreted and executed (ie. by turning ST into
LLVM and firing graalvm or c runtime or simply running it with one of
existing tools such as openplc), the remaining part will be possible
after proper adoption of code generated from MSpec.
Looking at amount of "soft plc" projects I have no doubts that its
possible. Its just matter of how much time and simplifications we can
accept.
I agree that we don't have a capacity nor will to dive into even more
specific parts of how PLCs work. If somebody else has such interest I
will definitely look for results and see if we could improve our part
afterwards.
Best,
Łukasz
On 4.05.2022 09:09, Christofer Dutz wrote:
Hi Lukasz,
I intentionally didn't mention the simulator, as I understood Peter, that he's
looking for something to run his PLC programs on for testing. Yes, the
simulator we have could be used to simulate being a PLC, but it can't actually
simulate the PLC itself. That's not the scope of that thing. I think
implementing the internal stuff needed to create a simulator where you can
deploy a program from TIA Portal is totally unthinkable, considering how many
different PLCs are out there.
Perhaps we should name it differently in order to manage expectations.
"plc-mock"?
And our simulator was actually created by me and it's up-to-date and we just
recently checked that it still works with the StreamPipes folks, as they were
looking for something to use in their StreamPipes integration-tests.
Chris
-----Original Message-----
From: Łukasz Dywicki <l...@code-house.org>
Sent: Dienstag, 3. Mai 2022 22:47
To: dev@plc4x.apache.org
Cc: pe...@petersaxton.uk
Subject: Re: Scope of using PLC4x as simulated driver for PLC projects
Hey Peter, welcome on mailing lists. First of all thank you for showing a
interest in a project and bringing interesting topic!
Just to oppose a little bit what Chris says, I would say there is no real
blocker for trying to do a PLC simulator within PLC4X. We do (or
did) have a S7 simulator in sandbox already. It was built a while ago by Julian
based on netty tcp server and I think 0.6 structures, so it might be a little
bit behind of what we currently have in a develop branch.
Technically speaking what you need in most of the cases is launching a TCP
server (in case of ADS/Modbus/S7) or a serial receiver (Modbus RTU/till some
degree CANopen) and answering requests which come over such communication
channel. If you get closer to basic fieldbus like a Modbus RTU or CANopen
difference between sending a request and answering it is not that large.
Few remarks - we currently do not have any support beside mentioned sandbox
project. If you decide to do such thing, it will be a far away from a realtime,
yet TCP/UDP stuff involving IP routing by definition is not! You will still
have to invest a ton of time in order to mimic a complete PLC behavior. I
suppose it might be problematic to keep it up with manufacturer tooling, simply
cause they can share some libraries or specifications unseen by any of us. Also
a device behavior can change with firmware upgrade, so you have to keep a close
look on that.
If you can sacrifice realtime requirement which is often crucial to
certification, live with above simplifications and caveats, then infrastructure
available in PLC4X will serve you better than any other open source components
available out there.
Before we go too optimistic, a fairly common problems is how devices react on
read/write operations. Each brand does it in an unique way, some models have
extended functionality, plus each device report errors differently. There is
also handling of parallel/overlapping requests which we handle partially even
in client side.
ADS for example has an internal notion of an routing table which is somehow
associated with communication peer, it has support for read single and read
multiple addresses, it has also support for subscriptions which you have to
coordinate. If you are not afraid of it, the communication structures will be
provided by MSpec. :)
I think that this is still a major advantage of MSpec approach and code
generation framework provided by PLC4X which awaits more use cases.
MSpec gives us a symmetric read/write operations which we prove through
serialization unit tests. This is unseen in most of libraries as they usually
focus on one of the sides - either a client/master or server/slave structures
and handle only half which is needed by them. A prime example could be a S7
client which does a request serialization and reply deserialization or S7
server which does request serialization and response serialization. Code
generated by PLC4X handles both ways.
Even if we miss a support for making server side framework, its something you
can definitely do on top of what PLC4X provide.
An interesting benefit you could get from relying on PLC4X is possibility to
make a test of different communications for assumed simulation logic. If you
find a way for example to write a PLC program in some portable way, then by
mapping such program inputs and outputs to PLC specific communication semantics
you get a very nice training suite.
If you search for ie. ANTRL IEC 61131-3 you will find a bunch of grammar files allowing
to parse structured text programs. From there you "just"
need to make interpreter working with one of languages supported by PLC4X code
generators.
Making a partial toolkit, even for a single brand/product line, is a lot of
work and code to be made. This code is potentially hard to maintain due to
various firmware versions, yet nothing really blocks you from trying. Do your
thing and report your findings if you wish. I am eager to hear your feedback. I
believe with your work we could improve our client side implementations too!
Cheers,
Łukasz
On 3.05.2022 13:55, Christofer Dutz wrote:
Hi Peter,
well simulating a real PLC generally would totally be out of scope of PLC4X. I
was more referring to the code to extract the program from a PLC and to deploy
it somewhere (like in a simulator or in case of Codesys you could run the
runtime on a test-server). But there are simulators out there. Usually, the PLC
vendors have them. Unfortunately, they usually are not for free. But I know
Siemens has various simulators to which you could deploy your program and then
interact with PLC4X.
Well, I guess you would need to find out how to make a software like TIA portal
believe your application is a S7 PLC and then need to implement the parts of
the protocol, that we haven't implemented yet. As S7 doesn't have a public
spec, you'll have to reverse engineer most of it. And another problem will be
that newer S7 devices use S7comm ... to convince TIA that you're a real device,
you would probably need to implement the S7comm protocol, which opens a whole
new box of worms.
All in all, I think the industry would need such a solution, but I also think
they haven't realized it yet and therefore are not willing to pay the money for
it.
Perhaps I'll try this concept out as part of my job, but that will probably
only have an effect on the types of PLCs that we use, and I have no idea, if
I'll get the time needed to do it.
Chris
-----Original Message-----
From: Peter Saxton <pe...@petersaxton.uk>
Sent: Dienstag, 3. Mai 2022 13:30
To: dev@plc4x.apache.org
Subject: RE: RE: Scope of using PLC4x as simulated driver for PLC
projects
Hi Chris,
Thanks for the clarification. So potentially PLC4X could drive the tests but
would leave the problem of simulating a PLC as outstanding.
One other approach I have head about is putting the PLC in a
completely virtual environment and then manipulating it by setting
values on the Virtual Machine. I don't have much information on this
approach just this blog post
https://www.plcnext-community.net/makersblog/painful-memories-of-plc-s
imulations/
You mentioned that you had though about working on something like this. What
would be the rough shape of your solution? Apart from industry resistance what
would be the hardest thing to get through in this project.
In my area there is a desire to embrace testing and CI probably not CD that
seems bold. At least the right noises are being made. We can see how it plays
out.
Peter
On 2022/05/03 08:17:49 Christofer Dutz wrote:
Hi Peter,
Well, if you want to use PLC4X to communicate with the real or the simulated
PLC in order to execute various test-scenarios: then probably PLC4X is what
you're looking for.
If you want to extract the PLC program from one PLC and write that to a
simulator, then unfortunately PLC4X doesn't have that focus and therefore
doesn't support this.
The programming part of automation engineering is what every PLC vendor
implements separately. Sometimes even one vendor has multiple ways of doing
that for their different products.
However, PLC4X does have the parts that you probably need to implement such a
feature, however it will certainly require reverse engineering for almost every
type of PLC.
I had been thinking about starting to work on something like what you describe.
However, my idea would have been to simulate just enough of a PLC for the
engineering tool to accept it and to have the Engineering tool deploy the new
program to that instead of a real PLC. This component would then pass the
program to a CI/CD pipeline, just as you describe it. However, I knew this
would be a lot of work and considering how well the industry accepted my offers
in the past, I decided that I won't do this, as it's not worth it. They haven't
even started thinking of concepts like continuous-integration or
continuous-delivery or even testing.
Chris
-----Original Message-----
From: Peter Saxton <pe...@petersaxton.uk>
Sent: Dienstag, 3. Mai 2022 09:35
To: dev@plc4x.apache.org
Subject: Scope of using PLC4x as simulated driver for PLC projects
Hi everyone.
First post here so apologies if my terminology is a bit off, I'm from a web
software background.
If I want to write behavior tests for a PLC project that already exists, is
PLC4x the right tool for that job or would there be something better?
Ideally I would be able to export a whole PLC project from a machine. Start
that exported project in a soft/simulated PLC environment and run a suite of
tests against the PLC logic. Ideally at no point would I be reliant on having
the correct hardware.
Reading through the PLC4x pages I found the section on "Testing (or using PLC4X
without a PLC)". As I understand this this is for testing software outside the PLC
but simulating a PLC. I would like to go the other way round and test a PLC project by
simulating the outside world. I also looked at the section on the simulated protocol, but
again I think this is for simulating values for a plc4x client.
Essentially we'd like to write a test saying "when input 1 & 2 are on the output A
is live", and check that the PLC code was going to do this.
I'm interested in testing against Siemens PLC projects, however for getting
started I would happily run a proof of concept on something like OpenPLC if
that was easier without hardware.
Cheers, Peter