EPICS: Allow reading partial waveforms (first N elements)

src/ophyd_async/epics/core/_aioca.py

@@ -294,8 +294,17 @@ async def connect(self, timeout: float):
         self.converter = make_converter(self.datatype, self.initial_values)
 
     async def _caget(self, pv: str, format: Format) -> AugmentedValue:
+        # Target: avoid breaking backwards compatability
+        #         perhaps overcautious on how to define element count
+        #         aioca.caget documentation states that element_count=1
+        #         by default. If this is provided values do change
+        # Todo:
+        #         need to check what pvaccess and p4p provide
+        kws = dict()
+        if self.options.element_count is not None:
+            kws["count"] = self.options.element_count
         return await caget(
-            pv, datatype=self.converter.read_dbr, format=format, timeout=None
+            pv, datatype=self.converter.read_dbr, format=format, timeout=None, **kws
         )
 
     def _make_reading(self, value: AugmentedValue) -> Reading[SignalDatatypeT]:

src/ophyd_async/epics/core/_signal.py

@@ -155,6 +155,7 @@ def epics_signal_r(
     read_pv: str,
     name: str = "",
     timeout: float = DEFAULT_TIMEOUT,
+    element_count: int | None = None
 ) -> SignalR[SignalDatatypeT]:
     """Create a `SignalR` backed by 1 EPICS PV.
 
@@ -163,7 +164,9 @@ def epics_signal_r(
     :param name: The name of the signal (defaults to empty string)
     :param timeout: A timeout to be used when reading (not connecting) this signal
     """
-    backend = _epics_signal_backend(datatype, read_pv, read_pv)
+    backend = _epics_signal_backend(
+        datatype, read_pv, read_pv, options=EpicsOptions(element_count=element_count)
+    )
     return SignalR(backend, name=name, timeout=timeout)
 
 

src/ophyd_async/epics/core/_util.py

@@ -38,6 +38,15 @@ class EpicsOptions(Generic[SignalDatatypeT]):
     as it causes a deadlock.
     """
 
+    element_count: None | int = None
+    """Epics allows to specify the maximum number of elements to transfer
+
+    Fast devices provide buffers that acquire large data sets. Typically
+    one only needs the beginning of this buffer.
+
+    None is used as standard argument: transfer as many elements as provided
+    """
+
 
 def get_pv_basename_and_field(pv: str) -> tuple[str, str | None]:
     """Split PV into record name and field."""