pyads-agile 0.3.0

Agile-maintained Python wrapper for the TwinCAT ADS library, based on pyads

pyads-agile

pyads-agile is a Python wrapper for the Beckhoff TwinCAT ADS library.

This distribution is maintained by Agile Automation Technologies GmbH and is based on the excellent upstream pyads project created by Stefan Lehmann: https://github.com/stlehmann/pyads

pyads-agile intentionally stays drop-in compatible with pyads. The public API, module name (import pyads), and supported interpreter/OS matrix mirror upstream, so existing applications can switch distributions without code changes. Current validated support target is Python 3.13 (CI runs on 3.13).

Attribution

• Original project: pyads by Stefan Lehmann
• Fork maintainer: Filippo Boido [e-mail] (Agile Automation Technologies GmbH)
• License: MIT
• This repository keeps upstream credit and license notices as required

See ACKNOWLEDGMENTS.md for details.

Installation

Install the distribution:

pip install pyads-agile

Import stays compatible:

import pyads

Versioning

pyads-agile uses its own independent Semantic Versioning (MAJOR.MINOR.PATCH). It does not mirror upstream pyads version numbers.

Scope

This package provides Python APIs for communicating with TwinCAT devices using:

• TcAdsDll.dll on Windows
• adslib.so on Linux

Agile-specific enhancements

Beyond compatibility, this fork currently focuses on improved RPC ergonomics:

• Convenient RPC object proxies. Connection.get_object() exposes TwinCAT function blocks as Python objects and lets you configure return and parameter types per method:

  TwinCAT requirement: each callable method must be annotated in PLC code with {attribute 'TcRpcEnable'} directly above the method declaration.

  rpc = plc.get_object(
      "GVL.fbTestRemoteMethodCall",
      method_return_types={"m_iSimpleCall": pyads.PLCTYPE_INT},
  )
  result = rpc.m_iSimpleCall()
  

• Multi-parameter RPC calls with native syntax. Configure method signatures once and then call methods like normal Python methods:

  rpc = plc.get_object(
      "GVL.fbTestRemoteMethodCall",
      method_return_types={"m_iSum": pyads.PLCTYPE_INT},
      method_parameters={"m_iSum": [pyads.PLCTYPE_INT, pyads.PLCTYPE_INT]},
  )
  result = rpc.m_iSum(5, 5)
  

• Typed RPC interfaces for IntelliSense. Decorate a Python class with @pyads.ads_path("GVL.fbTestRemoteMethodCall"), annotate method arguments and return types with TwinCAT PLC types, and pass the class into Connection.get_object. The returned proxy is typed as your class so IDEs can offer completions:

  @pyads.ads_path("GVL.fbTestRemoteMethodCall")
  class FB_TestRemoteMethodCall:
      def m_iSum(
          self,
          a: pyads.PLCTYPE_INT,
          b: pyads.PLCTYPE_INT,
      ) -> pyads.PLCTYPE_INT:
          ...
  
  rpc = plc.get_object(FB_TestRemoteMethodCall)
  result = rpc.m_iSum(5, 5)
  

  You can still use low-level direct calls when needed:

  result = plc.call_rpc_method(
      "GVL.fbTestRemoteMethodCall#m_iSimpleCall",
      return_type=pyads.PLCTYPE_INT,
      write_value=42,
      write_type=pyads.PLCTYPE_INT,
  )
  

• Serialized async ADS runtime. AsyncConnection executes all ADS calls on a dedicated worker thread per connection (in-order, race-safe on connection state), and exposes awaitable helpers and submit-style futures:

  import asyncio
  import pyads
  
  async def main() -> None:
      async with pyads.AsyncConnection("127.0.0.1.1.1", pyads.PORT_TC3PLC1) as plc:
          fut = plc.submit_sum_read(["GVL.int_val", "GVL.bool_val"])
          # ... do other work
          values = await fut
  
          await plc.sum_write({"GVL.int_val": int(values["GVL.int_val"]) + 1})
  
  asyncio.run(main())
  

• Async wrappers for Stefan's synchronous Connection API. AsyncConnection now mirrors the core synchronous read/write surface while keeping single-threaded serialized execution under the hood. For most methods you get both:

  • submit_* returning asyncio.Future
  • async method variant that awaits the same operation

  Covered wrappers include:

  • read, write, read_write
  • read_by_name, write_by_name
  • read_structure_by_name, write_structure_by_name
  • read_state, read_device_info, write_control
  • get_local_address, get_handle, release_handle, set_timeout
  • sum_read / sum_write (submit_sum_read / submit_sum_write)

  import asyncio
  import pyads
  
  async def main() -> None:
      async with pyads.AsyncConnection("127.0.0.1.1.1", pyads.PORT_TC3PLC1) as plc:
          # Await-style
          value = await plc.read_by_name("GVL.int_val", pyads.PLCTYPE_INT)
          await plc.write_by_name("GVL.int_val", value + 1, pyads.PLCTYPE_INT)
  
          # Submit-style
          fut = plc.submit_read_state()
          state = await fut
          print(state)
  
  asyncio.run(main())
  

• Async typed RPC objects. Use @pyads.ads_async_path(...) with AsyncConnection.get_async_object(...) for type-safe async RPC interfaces. Method calls return asyncio.Future objects:

  @pyads.ads_async_path("GVL.fbTestRemoteMethodCall")
  class FB_TestRemoteMethodCall:
      def m_iSum(
          self,
          a: pyads.PLCTYPE_INT,
          b: pyads.PLCTYPE_INT,
      ) -> asyncio.Future[pyads.PLCTYPE_INT]:
          ...
  
  async def main(plc: pyads.AsyncConnection) -> None:
      rpc = plc.get_async_object(FB_TestRemoteMethodCall)
      future = rpc.m_iSum(5, 5)
      result = await future
      print(result)
  

• Native stepchain async RPC interfaces. Use @pyads.ads_stepchain_path(...) to tell the async proxy that this is a long-running state-machine call. Calls return a StepChainOperation containing:

  • accepted: RPC-return phase
  • done: completion phase based on PLC status fields
  • await op: same as await op.done

  @pyads.ads_stepchain_path(
      "GVL.fbTestRemoteStepChainMethodCall",
      # Optional defaults shown explicitly:
      completion="poll",  # or "notify"
      status_field="stStepStatus",
      request_id_field="udiRequestId",
      request_id_arg="udiRequestId",
      busy_field="xBusy",
      done_field="xDone",
      error_field="xError",
      error_code_field="diErrorCode",
  )
  class FB_TestRemoteStepChainMethodCall:
      def m_xStartStepChain(
          self,
          udiRequestId: pyads.PLCTYPE_UDINT,
      ) -> pyads.PLCTYPE_BOOL:
          ...
  
  async def run_stepchain(plc: pyads.AsyncConnection) -> None:
      rpc = plc.get_async_object(FB_TestRemoteStepChainMethodCall)
  
      # udiRequestId is auto-generated if omitted.
      op = rpc.m_xStartStepChain()
  
      accepted = await op.accepted
      if not accepted:
          raise RuntimeError("Stepchain start rejected by PLC.")
  
      # Wait until status reports completion or error.
      await op
  
      # Built-in framework status read (predefined structure)
      status = await rpc.read_status()
      print(status["udiStep"], status["sStepName"])
  

  Completion backend options:

  • completion="poll": periodic sum_read checks (poll_interval/timeout_s)
  • completion="notify": ADS notifications trigger status reads in asyncio

  Built-in predefined stepchain status fields:

  • udiRequestId, xBusy, xDone, xError, diErrorCode, udiStep, sStepName

Features

• connect to remote TwinCAT devices
• create routes on Linux and on remote PLCs
• support for TwinCAT 2 and TwinCAT 3
• read and write values by name or by address
• read and write DUTs (structures)
• notification callbacks
• typed RPC interfaces via @pyads.ads_path(...)
• async typed RPC interfaces via @pyads.ads_async_path(...)
• serialized asyncio runtime via pyads.AsyncConnection
• async wrappers for core sync ADS methods (submit_* + awaitable variants)
• async typed RPC proxies via get_async_object(...)
• native stepchain async RPC flow via @pyads.ads_stepchain_path(...)
• stepchain completion backends: polling (poll) and notification-driven (notify)

Basic usage

import pyads

plc = pyads.Connection("127.0.0.1.1.1", pyads.PORT_TC3PLC1)
plc.open()
i = plc.read_by_name("GVL.int_val")
plc.write_by_name("GVL.int_val", i)
plc.close()

Contribution Policy

This repository is maintained on a best-effort basis for internal and product needs.

At this time, we do not accept unsolicited pull requests, and we may not be able to respond to feature requests or general support issues.