motivation: a thread creating a lot of messages like a polling loop
with very short polling frequency or a fast polling connection might
monopolize the output of message over receiving new messages.
In addition, the current design has a latency of 0.3 sec for the
output of asynchronous replies.
Anyway, the output queue is just extending the network output buffer,
which is usally big enough.
- change the name of 'queue_async_reply' to 'send_reply'.
This method anyway was not only used for async replies.
- send_reply is directly sending the reply instead of putting into the
queue. It will slow down the calling thread, if the output buffer
is full, which is desired behaviour.
Change-Id: I305669be2f7c027355b43421432f32be9c166ed4
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/23119
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
- send a heartbeat, if no events for 5 sec.
an interrupted connection (not closed by the other end) may
not be detected for a long time when nothing is sent
+ make the error reply on a non SECoPEror more verbose
e.g. "KeyError('foo')" instead of just "foo"
+ allow cfg file without nodeinterface
+ shorter logger name in HasIodev
Change-Id: I6b1ff23f9bf8c96feb25af44935596437b7d726f
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/23098
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
config file format change:
The section names no longer contain a space, the are either
bare module names or 'NODE' or 'INTERFACE' (capitalized in order to
distingish from module names).
The present code still accepts the old form.
Moving to the 'toml' format was considered too, but this needs some
more investigations. The necessary code changes would be limited
to the method Server.loadCfgFile.
Change-Id: I6020058c9dcc4c1cbf38f5b9e8f67e9aad670183
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/23031
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
in case a remote is disconnected and can be reconnected quickly,
a seamless routing is possible. However, when after reconnection
the description has changed, the server is restarted.
+ more information in error messages from dispatcher
Change-Id: I0837e0254aee3d12a26481f6fd697081a53aabba
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/22506
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
new client intended as base class for all clients
- based on Ennos secop client in nicos ([WIP] provide secop client)
- self healing connection singletons
- extension for other than TCP is foreseen (by extending new uri schemes)
- extensible name mangling
- seperate rx and tx threads supporting events
- internal cache
- extensible error handling
- callback for unhandled messages
- callback for descriptive data change
- callback for node stat change (connected, disconnected)
- a short close down and reconnect without change in descriptive data
does not disturb the client side
works with secop-gui (change follows), planned to be used for Frappy
internal secop proxy and as a replacement for secop.client.baseclient.Client
in the nicos secop device.
-> secop/client/baseclient.py to be removed after planned changes
moved secop/client/__init__.py to secop/client/console.py because secop.client
would be the natural place to put the new base class.
Change-Id: I1a7b1f1ded2221a8f9fcdd52f9cc7414e8fbe035
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/22218
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
modules using the old poller (now called secop.poller.BasicPoller)
need to explcitly declare it (pollerClass = BasicPoller)
BasicPoller is still used in:
secop_mlz/amagnet.py
secop_mlz/entangle.py
secop/simulation.py
Remark: before removing BasicPoller we may need a replacement for
Readable.pollParams
Change-Id: If1ae8b68e02f13e601334656b818337c882e06cc
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/21910
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
this change is triggered by the fact, that assigining a unit
in the config file did no longer work.
this change has several implications:
1) secop.properties must not import secop.datatypes:
- as ValueType can not be imported, the default behaviour with
'mandatory' and 'default' arguments was slightly changed
- instead of checking for DataType when exporting, a try/except
was used
2) the datatype of datatype properties is sometimes not yet defined.
a stub is used in this cases instead, which is later replaced by
the proper datatype. The number of stubs may be reduced, but this
should be done in a later change, as the diff will be much less
readable.
3) in config files, datatype properties can be changed like parameter
properties. HasProperties.setProperties/checkProperties/getProperties
are overridden for this.
the config editor seems still to work, an issue (probably py3) had
to be fixed there
Change-Id: I1efddf51f2c760510e913dbcaa099e8a89c9cab5
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/21399
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
- None <= 0 is invalid in py3
- restrict BLOBType to bytes (may be changed after migration to py3)
- remove long
- use list(<dict>.items()) when dict is changed within loop
- allow initialization of properties in HasProperties without
supercall to base class
- <dict>.values() can not be indexed
- adapted/removed various tests. additional tests might be added
after definitive migration to py3
after this change, all the tests run with py3, also secop-server and
secop-gui were tested with an example, but other code might still contain
py3 incompatibilities
Change-Id: I881c6972aeabb8494a21a6cbc7ffeddfd4f5d4f8
Reviewed-on: https://forge.frm2.tum.de/review/c/sine2020/secop/playground/+/21306
Tested-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
Reviewed-by: Bjoern Pedersen <bjoern.pedersen@frm2.tum.de>
messages might get lost after activate, when the number of parameters
exceeds maxlen of the output queue, as a limited deque silently overwrites
messages.
first try: use an infinite deque, but slow down appending thread when deque
gets too big
-> does not work because the appending thread is the same
second try: increase maximum a lot, on overflow log an error message and
close connection
Change-Id: I20376f7e08240dabe43269fa63c596f07e59ddf6
Reviewed-on: https://forge.frm2.tum.de/review/20982
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
+ DataType validators are shifted to __call__
+ as_json is moved to export_datatape()
+ new HasProperties Base Mixin for Modules/DataTypes
+ accessibles can be accessed via iterator of a module
+ properties are properly 'derived' and checked, are set with .setPropertyValue
remember: parameters only have properties, so use getPropertyValue()
Change-Id: Iae0273f971aacb00fe6bf05e6a4d24a6d1be881a
Reviewed-on: https://forge.frm2.tum.de/review/20635
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
needs a bigger rework, since READREPLY and EVENTREPLY are now different....
Also the format of the error-reply got changed :(
Change-Id: I1760743238227730ee49aaf92b54e0ff5f25423b
Reviewed-on: https://forge.frm2.tum.de/review/20246
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
- unify (encoded) Errormessages. They now always contain the offeding request.
- adopt keywords to current spec
- fixed formatting of timestamp in WriteReply
- minor bugfixing + polishing
Change-Id: I0ab8836756551399643bdce3d062eedd345509f1
