Changelog

Structured data exchange and JSON support for C/C++

All notable changes to the Sigmyne/xchange library will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Upcoming feature release, possibly as early as 1 August 2026.

CMake xchangeConfig to skip requiring math lib for non-Windows platforms in general, since it can fail if the math library is in the build path, but not in the search path, such as for some cross builds (see e.g. the vcpkg Android builds)

#32: Added XBoolean as the xchange-specific boolean type, replacing boolean in prior versions. (The boolean type is still defined for compatibility, but can be disabled by defining _TYPEDEF_BOOLEAN prior to including xchange.h in your source code.).
Added xIsDebug() function to check on xDebug. While the global variable is fine in most cases, they are problematic for Windows DLLs. It’s better to use purely functional access instead.
xvprintf() / xdprintf() macros to reference functions instead of global vars (see above comment on Windows DLLs).

#32: Changed boolean parameter and return types to the equivalent XBoolean, to disambiguate the xchange-specific boolean type from other definitions. (Other header may define boolean also, and so by choosing a more unique type name, we reduce the chance of namespace conflicts.)

High priority bug fixes and version bump.

#26: Fixed space allocation in xLookupPutAllAsync() / xLookupRemoveAllAsync().
#28: xGetAsDoubleAtIndex() returned integer rounded values when the field stored float or double type data. Now, it returns the floating-point value directly.
Fixed memleak in xLookupRemove().
Fixed memleak in xDestroyLookup().
xLookupPut() did not return X_NO_INIT as expected if lookup table was not initialized.
Fix potential buffer overflow at build time in docedit.c (sprintf() to snprintf()).
Fixed max string size for 8-byte signed integer (-> 20 bytes + termination).
Fixed insufficient checking in xIsFieldValid().
Fixed botched read loop in xjsonParseFile().
Removed the unwanted fclose() in xjsonParseFile(). Caller is responsible for closing the file after the call, as appropriate.
Extra comma when printing X_FLOAT value to JSON.
Fixed JSON unicode processing.
Fixed PrintPrimitive() return value for X_CHAR(n) types in xjson.c.
xGetAsLongAtIndex() and xGetAsDoubleAtIndex() for string/raw types.

#29: Use snprintf() instead of sprintf() provided it’s available. (On older platforms prior to the C99 standard, it defaults to sprintf().)
#29: Use size_t or long (if needs to be signed) instead of int for string length parameters.
Changed xlookup hash algorithm to FNV-1a.
Rounding instead of downcast from double in xGetAsLongAtIndex().
CMake install to skip .gitignore in examples/.
Makefile doc install to match CMake.
CMake export targets from build-directory (for dependent builds)
examples/Makefile to work standalone, without config.mk.

Maintenance release with improved portability (esp. Windows, MacOS, and BSD). It also enables ports to vcpkg and Homebrew (coming very soon…).

#23: Removed unneeded includes of UNIX-specific unistd.h (blocking portability).

#23: Use portable mutexes in xlookup.c.
#23: xjson error/warning reporting via static functions instead of niche macros that Windows does not support.
#23: xGetAsLongAtIndex() and xGetAsDoubleAtIndex() switch to setting errno to EINVAL instead of ENOSR, as the latter is not defined on BSD.

Maintenance release with minor code style improvements.

Minor feature release with bug fixes.

#15: xPrintFloat() printed an extra digit, which would appear as a ‘rounding error’ in decimal representations.
#16: Width detection of platform-specific built-in integer types (i.e., short, int, long, and long long). The previous implementation included stdint.h with __STDC_LIMIT_MACROS defined. However, if the application source, then included stdint.h before xchange.h, then the fixed-width integer limits were left undefined. As a result, we no longer rely on stdint.h providing these limits.

xParseFloat() to parse floats without rounding errors that might result if parsing as double and then casting as float.

Bug fix release.

xDestroyLookupAndData() to destroy a lookup table including all the data that was referenced inside it.

xCreateField() to treat X_RAW types always as scalars, ignoring the dimensions provided.

Initial public release.