All notable changes to this project will be documented in this file. The format is based on Keep a Changelog and this project adheres to Semantic Versioning.
- {cmake} Generate a package version file (#316) (Thanks SunBlack!)
- {cmake} Require XercesC 3.2 and remove
USING_STATIC_XERCES
(#317) (Thanks SunBlack!)
-
{cmake} Added missing include for GNUInstallDirs (#315) (Thanks Adrian!)
If you built without testing on, the cmake files were not installed to the correct location.
3.2.0 - 2024-06-27
-
{cmake} Added
E57_INSTALL_CMAKEDIR
option to override where the cmake files are installed. (#305)This defaults to
lib/cmake/E57Format
and the path is relative toCMAKE_INSTALL_PREFIX
.
- {cmake} If building as a shared library, set VERSION & SOVERSION. (#304)
- {cmake} Don't force warnings as errors when building self. (#299)
- {cmake} Use git tag in library version for more precision. (#298)
- {license} Re-license all my (asmaloney) MIT files to BSL-1.0. It is just easier to have it all under one license. (#296)
- {format} Update to clang-format 18 & reformat code. (#286)
- Add "E57_" to macros in E57Exception.h. (#285)
- "De-deprecate" methods in E57SimpleWriter. These methods can be useful when writing batches. (#284)
- Update CRCpp to fix MSVC and MinGW warnings. (#302)
- Fix warnings about redefinition of
_LARGEFILE64_SOURCE
and__LARGE64_FILES
(#301) (Thanks Niklas!) - Fix building with emscripten. (Note that the project doesn't officially support emscripten.) (#288) (Thanks maz-1!)
- {standard conformance} CompressedVectors always write an index packet. This is required by the standard (9.3.5). (#295)
- {standard conformance} E57SimpleReader accepts files containing zero scans. (#283)
- {cmake} Replace deprecated "exec_program" with "execute_process". (#282)
- Fix potential invalid range exceptions when reading integer nodes. (#278)
3.1.1 - 2023-11-28
- Fix library version in CMakeLists.txt. (Thanks Olli!)
3.1.0 - 2023-11-20
-
{cmake} New option
E57_RELEASE_LTO
controls whether link-time optimization is on for release builds. It defaults toON
. (#254)CMake forces "thin" LTO (see this issue) which is a problem if compiling statically for distribution (e.g. in a package manager). Generally you will only want to turn this off for distributing static release builds.
-
Trying to read a Data3D with zero records which has an ill-formed header will now throw an
ErrorData3DReadInvalidZeroRecords
exception instead ofErrorInternal
. (#264)Note that previous versions of this library (and E57RefImpl itself) could write these headers incorrectly. This was also fixed in this release (see below).
-
{cmake} Remove
E57_VISIBILITY_HIDDEN
option. (#259)I cannot get extern templates to work across all of gcc, clang, apple clang, and MSVC when using "hidden" visibility with shared libraries.
- Fix #include to avoid MSVC compilation error with Visual Studio 2017 (#268) (Thanks Thomas!)
- {standard conformance} A compressed vector with 0 records must still write a data packet. (#266)
- {standard conformance} Fix reading a compressed vector with 0 points which has an empty data packet. (#267)
- {standard conformance} Compressed vectors with an invalid section ID now throw an
ErrorBadCVPacket
exception ifE57_VALIDATION_LEVEL
> 0. (#265) - Fix clang warnings about implicit conversions in SourceDestBufferImpl.cpp. Apple's clang doesn't warn about these, but it looks like the official clang releases do. (#257) (Thanks Martin!)
3.0.2 - 2023-07-22
- Fix
E57_VERBOSE
build. (#251) (Thanks Jia!) - {standard conformance} Fix invalid range exception in FloatNode implementation. (#250)
- Fix reading of index packets. (#249)
- Fix several places where we should be checking for MSVC, not WIN32. (#248)
- Fix "unnecessary semicolons" warnings which prevented building with GCC <= 10. (#241) (Thanks Andre!)
3.0.1 - 2023-03-15
- {ci} Added an MSVC 32-bit build. (#235)
- {cmake} Turn off inter-procedural optimization in debug builds. Link-time optimization can make debugging more difficult. (#240)
- {cmake} Don't force a debug postfix if
CMAKE_DEBUG_POSTFIX
is defined.(#239) - {cmake} Don't force install locations. This prevents overriding them using
CMAKE_INSTALL_BINDIR
&CMAKE_INSTALL_LIBDIR
.(#237) - Fix warnings which prevented building on 32-bit systems. (#233, #234)
3.0.0 - 2023-02-23
There have been many changes around the Simple API
in this release to fix some problems, avoid potential errors, and simplify the use of the API. Where possible these changes are marked as deprecated
to be removed later. The compiler will produce warnings for these indicating what needs to change. Other changes could not be marked deprecated but will break the API, requiring code changes. The notable ones are marked with 🚧 below.
-
Add address (ASan) and undefined behaviour (UBSan) sanitizers. These are controlled by
E57FORMAT_SANITIZE_ALL
,E57FORMAT_SANITIZE_ADDRESS
andE57FORMAT_SANITIZE_ADDRESS
. They are not included when building with MSVC. -
Add new
E57Version.h
header for more convenient access to version information. (#197). -
Add
ImageFile::extensionsLookupPrefix()
overload for more concise user code. (#161) -
Added a constructor & destructor for E57SimpleData's
Data3DPointsData_t
. This will create all the required buffers based on ane57::Data3D
struct and handle their cleanup. See theSimpleWriter
tests for examples. (#149)Note: I strongly recommend these new constructors be used to simplify your code and help prevent errors.
-
A new E57SimpleReader constructor takes a
ReaderOptions
struct which allows setting the checksum policy.Reader( const ustring &filePath, const ReaderOptions &options );
-
{test} Added testing using GoogleTest. For details, please see test/README.md (#121)
-
Added
E57Exception::errorStr()
to get the error string directly. (#128) -
{ci} Added a CI check for proper clang-formatted code. (#125)
- Now requires a C++14 compatible compiler.
- Now requires CMake >= 3.15. (#205)
- 🚧 DEBUG and VERBOSE macros were changed to simplify and clarify:
- New
E57_ENABLE_DIAGNOSTIC_OUTPUT
controls the inclusion of the code for thedump()
functions on nodes. I don't see any real reason to turn this off, but I left the capability in for compatibility. - New
E57_VALIDATION_LEVEL=N
replacesE57_DEBUG
(N=1) andE57_MAX_DEBUG
(N=2). E57_MAX_VERBOSE
was consolidated withE57_VERBOSE
since they were essentially the same.
- New
- When building itself, warnings are now treated as errors. (#205, #211)
- Clean up global const and enum names to use the
e57
namespace and avoid repetition. (#176)- i.e. instead of
e57::E57_STRUCTURE
, we now usee57::TypeStructure
- i.e. instead of
- {format} Update clang-format rules for clang-format 15. (#168, #179)
- Change default checksum policies to an enum. (#166)
Old New CHECKSUM_POLICY_NONE ChecksumPolicy::None CHECKSUM_POLICY_SPARSE ChecksumPolicy::Sparse CHECKSUM_POLICY_HALF ChecksumPolicy::Half CHECKSUM_POLICY_ALL ChecksumPolicy::All - Avoid implicit conversion in constructors. (#135)
- Update CRCpp to 1.2. (#130)
- E57Exception changes (#118):
- mark methods as
noexcept
- use
private
instead ofprotected
- mark methods as
- Rename E57Simple's
Data3DPointsData
andData3DPointsData_d
structs toData3DPointsFloat
andData3DPointsDouble
respectively. (#180) - 🚧 E57Simple: Specifying the node type for cartesian & spherical points, time stamp, and intensity is now explicit using new fields (
pointRangeNodeType
,angleNodeType
,timeNodeType
, andintensityNodeType
) and a new enum (NumericalNodeType
). (#178)- For examples, please see test/src/testSimpleWriter.cpp.
- Simplify the E57SimpleWriter API with
WriteImage2DData()
for images andWriteData3DData()
for 3D data. This reduces code, hides complexity, and avoids potential errors. (#171)- As part of this simplification,
WriteData3DData()
will now fill in any missing min/max values for cartesian & spherical points, intensity, and time stamps by looking at the data.(#175)
- As part of this simplification,
- 🚧 E57Simple: Intensity now uses
double
instead offloat
. (#178) - 🚧 E57Simple: Colours now use
uint16_t
instead ofuint8_t
. (#167) - 🚧 Change E57SimpleData's Data3D field name from
pointsSize
topointCount
. (#164) - 🚧 Min/max fields in E57SimpleData's Data3D's point fields were a mix of floats and doubles. Since the fields are doubles, set them all to doubles and use the new
Data3DPointsData_t
constructor to set them properly for floats. (#153)Note: If you were previously relying on these to be floats and are not using the new
Data3DPointsData_t
constructor, you will need to set these. - 🚧 Renamed the E57_EXT_surface_normals extension's fields in E57SimpleData's
PointStandardizedFieldsAvailable
to be in line with existing code. (#149)normalX
renamed tonormalXField
normalY
renamed tonormalYField
normalZ
renamed tonormalZField
e57::Utilities::getVersions()
. (#197)e57::Data3DPointsData
ande57::Data3DPointsData_d
types. (#180)- Many global const and enum names. The compiler will produce warnings including the replacement symbols (note that enumerators will not produce warnings on C++14, but they will on C++17). (#176)
e57::Writer::NewImage2D
ande57::Writer::SetUpData3DPointsData
. (#171)- Old default checksum policies (
CHECKSUM_POLICY_NONE
,CHECKSUM_POLICY_SPARSE
,CHECKSUM_POLICY_HALF
, andCHECKSUM_POLICY_ALL
). (#166) - The old
e57::Reader
constructor taking onlyfilePath
. (#139) - The old
e57::Writer
constructor taking onlyfilePath
. (#117)
- Fix several potential divide-by-zero cases. (#224)
- {ci} Now builds shared library versions as well. (#219)
- {win} Fixes shared library build warnings. (#215, #216, #217)
- Fix the code which shortens floating point numbers when converted to strings. The impact of it being incorrect was negligible since it's just the floating point representation in the XML portion of the file. (#214)
- Turned on a lot of compiler warnings and fixed them. (#201, #202, #203, #204, #205, #207, #209)
- Fix writing floating point numbers when
std::locale::global
is set. (#174) - E57XmlParser: Parse doubles in a locale-independent way. (#173) (Thanks Hugal31!)
- E57SimpleReader: Ensure scaled integer fields are set as best we can when reading. (#158)
- Fix the E57_EXT_surface_normals extension's URI in E57SimpleWriter. (#143)
- {win} Fix conversion warning when compiling with debug on. (#124)
- Add errno detail to
E57_ERROR_OPEN_FAILED
exception. (#119, #120)
2.3.0 - 2022-10-04
- {cmake} Added
E57_VISIBILITY_HIDDEN
option to control CXX_VISIBILITY_PRESET. Defaults toON
. (#104) (Thanks Nigel!) - Added BSD support. (#99) (Thanks Christophe!)
- Updated & reorganized the online API docs and changed to a cleaner theme.
- Change file creation to use 0666 permissions on POSIX systems. (#105) (Thanks Nigel!)
- {cmake} CXX_VISIBILITY_PRESET is now set and
ON
by default. (#104) (Thanks Nigel!) - A new E57SimpleWriter constructor takes a
WriterOptions
struct which allows setting the file's GUID.The old constructor taking onlyWriter( const ustring &filePath, const WriterOptions &options );
coordinateMetadata
is deprecated and will be removed in the future. (#96) (Thanks Nigel!) - Change
E57_DEBUG
,E57_MAX_DEBUG
,E57_VERBOSE
,E57_MAX_VERBOSE
,E57_WRITE_CRAZY_PACKET_MODE
from #defines to cmake options. (#80) (Thanks Nigel!)
- Fix E57SimpleWriter's writing of invalid quaternions. It now defaults to the identity quaternion. (#108) (Thanks Nigel!)
- Fix E57SimpleReader to handle missing
images2D
andisAtomicClockReferenced
nodes. (#90) (Thanks Olli!) - Fix BitpackIntegerDecoder sometimes reading past end of input buffer. (#87) (Thanks Nigel!)
- Fix compilation when some debug options are set. (#81, #82, #84) (Thanks Nigel!)
- Fix compilation with musl libc (#70) (Thanks Dimitri!)
- Add missing include for GCC 11 (#68) (Thanks bartoszek!)
Note: The next release will be 3.0 and will require a C++14 compiler.
2.2.0 - 2021-04-01
- Added and updated the E57Simple API from the old reference library. (#41, #63) (Thanks Jiri & Grégoire!)
- Enabled building E57Format as a shared library. (#40) (Thanks Amodio!)
- Added a clang-format file, a cmake target for it ("format"), and reformatted the code.
- {doc} Added info about using SPDX License Identifiers.
- {ci} Added GitHub Actions to build macOS, Linux, and Windows. (#35)
E57_V1_0_URI
was changed from a#define
to aconstexpr
, so if you use it, it will need to be updated with a namespace:e57::E57_V1_0_URI
.- {doc} Moved some documentation to new repo (libE57Format-docs) and generate the docs.
- {cmake} Reviewed and updated cmake files. CMake minimum version was changed to 3.10.
- Removed all internal usage of dynamic_cast<>. (#39) (Thanks Jiri!)
- Split classes out from E57FormatImpl.[h,cpp] into their own files.
2.1.0 - 2020-04-01
- Added support for UTF8 file names on Windows. (based on #26)
- Added support for char* input. (#22)
- {cmake} Added fallback configuration for RelWithDebInfo and MinSizeRel. (#29)
- {cmake} Added a proper install configuration. (#28)
- {cmake} Removed unused ICU requirement for Linux.
- {cmake} Marked xerces-c as required.
- {cmake} Various cleanups.
- Internal code cleanups.
2.0.1 - 2019-01-15
- Writing files was broken and would produce the following error:
Error: bad API function argument provided by user (E57_ERROR_BAD_API_ARGUMENT) (ImageFileImpl.cpp line 109)
2.0.0 - 2019-01-06
Forked from E57RefImpl.
- Added a checksum policy (see ReadChecksumPolicy in E57Format.h) so the library user can decide how frequently to check them.
- {win} Added cmake option ()
USING_STATIC_XERCES
) to tell the build if you are using a static Xerces lib.
- Now requires C++11.
- Now requires cmake 3.1+.
- No longer uses BOOST.
- Turn off
E57_MAX_DEBUG
by default.
- Removed all but the main sources for reading and writing E57 files.
- Removed "big endian" byte swap code (not sure it was working and no way to test).
- Multiple fixes for compilation on macOS.
- Fixed a couple of fallthrough bugs which would result in undefined behaviour.
- Improved file read times.
- Many, many code cleanups:
- Refactored the code into multiple files.
- Removed unused macros and code.
- Removed non-useful comments.
- Added proper initialization of class and struct members.
- Modernized using c++11.