|
PVData C++
8.0.6
|
All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Groups Pages
|
PVData C++
8.0.6
|
The changes since release 6.0.0 are:
Fix "Problem building pvDataCPP for win32-x86-mingw" (issue #42) In src/misc/bitSet.cpp #include "algorithm" required for MSVS 2015 In testApp/misc/testTypeCast.cpp print (u)int8 values as integers Minor documentation updates
The main changes since release 5.0.4 are:
Linux shared library version added Headers have been moved into pv directories Bitset functions declared const where possible Bitset::swap added Requester::message has default implementation Serialization/deserialization helpers added Non-template getSubField char* overload added MonitorPlugin deprecated Field name validation performed Now builds for Cygwin and MinGW targets Fix for debug build issue. New license file replaces LICENSE and COPYRIGHT
Linux shared library version numbers have been added by setting SHRLIB_VERSION (to 6.0 in this case). So shared object will be libpvData.so.6.0 instead of libpvData.so.
E.g. src/property/alarm.h -> src/property/pv/alarm.h
This facilitates using some IDEs such as Qt Creator.
Requester::message is no longer pure virtual. Default implementation sends string to std::cerr.
A helper function, serializeToVector, has been added which serializes a Serializable object into a standard vector of UInt8s.
Similarly a function deserializeFromVector deserializes a standard vector into a Deserializable object.
A function deserializeFromBuffer deserializes a ByteBuffer into a Deserializable object.
On creating a Structure or Union the field names are now validated.
Valid characters for a field name are upper or lowercase letters, numbers and underscores and intial numbers are invalid, i.e. names must be of the form [A-Za-z_][A-Za-z0-9_]*.
Includes cross-compiling MinGW on Linux.
The changes since release 5.0.3 are:
Fixed bitset serialization (issue #24) Fixed truncation in BitSet::or_and (issue #27)
C++ bitset serialization was not consistent with the C++ deserialization and Java code in some instances (depending on the endianness of the serializer and deserializer) when the number of bits was 56-63 modulo 64. C++ serialization has been fixed.
Fix exposed issue in deserialization on 32-bit platforms which has also been corrected.
If n, n1 and n2 words are used to store the values of the bitsets bitset, bitset1 and bitset2 respectively then max(n, min(n1,n2)) words are needed to store bitset.or_(bitset1, bitset2).
Previously min(n1,n2) words were used and the result would be truncated in some instances. This has been fixed.
The only change since release 5.0.2 is:
A PVUnion whose stored value was null was serialized without checking whether the buffer had sufficient capacity. This has been fixed by calling ensureBuffer().
The main changes since release 4.0 are:
The following methods have been removed from PVStructure
Use template getSubField instead, e.g. use
getSubField< PVInt >(fieldName)
in place of
getIntField(fieldName)
Convert copy methods where moved and replaced with methods on PVField classes, i.e.
PVField::copy(const PVField& from)
Methods
PVField::copyUnchecked(const PVField& from)
were added to allow unchecked copies, to gain performance where checked are not needed (anymore).
In addition:
Before this method, depending on types for to and from, sometimes did a shallow copy, i.e. just made to shared_ptr for to share the same data as from. Now it always copies between the subfield of to and from.
PVStructure has a new template member
getSubFieldT(std::string const &fieldName)
that is like getSubField() except that it throws a runtime_error instead of returning null.
This was mainly used in the implementation of getSubField. With a change to the latter, findSubField was removed.
New steam operators are available for Field and PVField. Before to print a Field (or any extension) or a PVField (or any extension) it was necessary to have code like:
void print(StructureConstPtr struc, PVStructurePtr pv)
{
if(struc) {
cout << *struc << endl;
} else {
cout << "nullptr\n"
}
if(pv) {
cout << *.struc << endl;
} else {
cout << "nullptr\n"
}
}
Now it can be done as follows:
void print(StructureConstPtr struc, PVStructurePtr pv)
{
cout << struc << endl;
cout << pv << endl;
}
A new template getField method has been added to Structure
template<typename FT > std::tr1::shared_ptr< const FT > getField(std::string const &fieldName) const
Can be used, for example, as follows:
StructurePtr tsStruc = struc->getField<Structure>("timeStamp");
Certain static builds (in particular Windows builds) of applications using pvData had issues due to PVStructure::DEFAULT_ID being used before being initialised. This has been fixed.
createRequest could cause a SEGFAULT if passed a bad argument. This has been changed so the it returns a null pvStructure and provides an error.
The main changes since release 3.0.2 are:
PVScalarArray, PVStructureArray, and PVUnionArray all enforce COW (Copy On Write) Semantics. In order to limit memory usage the storage for raw data is managed via a new shared_vector facility. This allows multiple instances of array data to use the shared raw data. COW is implemented via shared_vectors of const data, i. e. data that can not be modified.
This is replaced by std::string.
In timeStamp nanoSeconds is changed to nanoseconds.
In valueAlarm hysteresis is changed to hysteresis
pvData.h and pvIntrospect no longer defines toString Instead they have stream support. pvIntrospect uses method dump and pvData uses dumpValue. For example:
PVDoublePtr pvValue; String buffer; pvValue->toString(&buffer); cout << buffer << endl; buffer.clear(); pvValue->getField()->toString(&buffer); cout << buffer << evdl;
is replaced by
PVDoublePtr pvValue; cout << *pvValue << endl cout << *pvValue->getField() << endl;
There are two new basic types: union_t and unionArray.
A union is like a structure that has a single subfield. There are two flavors:
The field type can be dynamically changed.
This consists of createRequest and pvCopy. createRequest was moved from pvAccess to here. pvCopy is moved from pvDatabaseCPP and now depends only on pvData, i.e. it no longer has any knowledge of PVRecord.
This is for is for use by code that implements pvAccess monitors. This is prototype and is subject to debate.
This was the starting point for RELEASE_NOTES
1.8.5