- Making ICU Smaller
- ICU Versions
- Packaging ICU4C as Part of the Operating System
This chapter describes, for the advanced user, how to package ICU4C for distribution, whether alone, as part of an application, or as part of the operating system.
There are many ways that a person can package ICU with their software products. Usually only the libraries need to be considered for packaging.
On UNIX, you should use
gmake install to make it easier to develop and package ICU. The bin, lib and include directories are needed to develop applications that use ICU. These directories will be created relative to the
--prefix=dir” configure option (See the UNIX build instructions). When ICU is built on Windows, a similar directory structure is built.
When changes have been made to the standard ICU distribution, it is recommended that at least one of the following guidelines be followed for special packaging.
- Add a suffix name to the library names. This can be done with the
- The installation script should install the ICU libraries into the application’s directory.
Following these guidelines prevents other applications that use a standard ICU distribution from conflicting with any libraries that you need. On operating systems that do not have a standard C++ ABI (name mangling) for compilers, it is recommended to do this special packaging anyway. More details on customizing ICU are available in the User Guide. The ICU Source Code Organization section of the ICU4C gives a more complete description of the libraries.
ICU has several libraries for you to use. Here is an example of libraries that are frequently packaged.
|Library Name||Windows Filename||Linux Filename||Comment|
|Data Library||icudtXYl.dll||libicudata.so.XY.Z||Data required by the Common and I18n libraries. There are many ways to package and customize this data, but by default this is all you need.|
|Common Library||icuucXY.dll||libicuuc.so.XY.Z||Base library required by all other ICU libraries.|
|Internationalization (i18n) Library||icuinXY.dll||libicui18n.so.XY.Z||A library that contains many locale based internationalization (i18n) functions.|
|Layout Extensions Engine||iculxXY.dll||libiculx.so.XY.Z||An optional engine for doing paragraph layout that uses parts of ICU. HarfBuzz is required.|
|ICU I/O (Unicode stdio) Library||icuioXY.dll||libicuio.so.XY.Z||An optional library that provides a stdio like API with Unicode support.|
|Tool Utility Library||icutuXY.dll||libicutu.so.XY.Z||An internal library that contains internal APIs that are only used by ICU’s tools. If you do not use ICU’s tools, you do not need this library.|
Normally only the above ICU libraries need to be considered for packaging. The versionless symbolic links to these libraries are only needed for easier development. The X, Y and Z parts of the name are the version numbers of ICU. For example, ICU 2.0.2 would have the name libicuuc.so.20.2 for the common library. The exact format of the library names can vary between platforms due to how each platform can handles library versioning.
The ICU project is intended to provide everything an application might need in order to process Unicode. However, in doing so, the results may become quite large on disk. A default build of ICU normally results in over 16 MB of data, and a substantial amount of object code. This section describes some techniques to reduce the size of ICU to only the items which are required for your application.
If you add the
--enable-static option to the ICU command line build (Makefile or cygwin), ICU will also build a static library version which you can link to only the exact functions your application needs. Users of your ICU must compile with -DU_STATIC_IMPLEMENTATION. Also see How To Use ICU.
ICU consists of a number of different libraries. The library dependency chart in the Design chapter can be used to understand and determine the exact set of libraries needed.
Certain features of ICU may be turned on and off through preprocessor defines. These switches are located in the file “uconfig.h”, and disable the code for certain features from being built.
All of these switches are defined to ‘0’ by default, unless overridden by the build environment, or by modifying uconfig.h itself.
|Switch Name||Library||Effect if #defined to ‘1’|
|UCONFIG_ONLY_COLLATION||common & i18n||Turn off all other modules named here except collation and legacy conversion|
|UCONFIG_NO_LEGACY_CONVERSION||common||Turn off conversion apart from UTF, CESU-8, SCSU, BOCU-1, US-ASCII, and ISO-8859-1. Not possible to turn off legacy conversion on EBCDIC platforms.|
|UCONFIG_NO_BREAK_ITERATION||common||Turn off break iteration|
|UCONFIG_NO_COLLATION||i18n||Turn off collation and collation-based string search.|
|UCONFIG_NO_FORMATTING||i18n||Turn off all formatting (date, time, number, etc), and calendar/timezone services.|
|UCONFIG_NO_TRANSLITERATION||i18n||Turn off script-to-script transliteration|
|UCONFIG_NO_REGULAR_EXPRESSIONS||i18n||Turn off the regular expression functionality|
NOTE: These switches do not necessarily disable data generation. For example, disabling formatting does not prevent formatting data from being built into the resource bundles. See the section on ICU data, for information on changing data packaging. However, some ICU data builders will not function with these switches set, such as UCONFIG_NO_FILE_IO or UCONFIG_NO_REGULAR_EXPRESSIONS. If using these switches, it is best to use pre-built ICU data, such as is the default for ICU source downloads, as opposed to data builds “from scratch” out of SVN.
This method involves setting an environment variable when ICU is built. For example, on a POSIX-like platform, settings may be chosen at the point runConfigureICU is run:
env CPPFLAGS="-DUCONFIG_NO_COLLATION=1 -DUCONFIGU_NO_FORMATTING=1" \ runConfigureICU SOLARISCC ...
Note: When end-user code is compiled, it must also have the same CPPFLAGS set, or else calling some functions may result in a link failure.
This method involves modifying the source file icu/source/common/unicode/uconfig.h directly, before ICU is built. It has the advantage that the configuration change is propagated to all clients who compile against this build of ICU, however the altered file must be tracked when the next version of ICU is installed.
Modify ‘uconfig.h’ to add the following lines before the first #ifndef UCONFIG_… section
#ifndef UCONFIG_NO_COLLATION #define UCONFIG_NO_COLLATION 1 #enddif #ifndef UCONFIG_NO_FORMATTING #define UCONFIG_NO_FORMATTING 1 #endif
There are many ways in which ICU data may be reduced. If only certain locales or converters will be used, others may be removed. Additionally, data may be packaged as individual files or interchangeable archives (.dat files), allowing data to be installed and removed without rebuilding ICU. For details, see the ICU Data chapter.
The following table gives an example of the dynamically linked library and symbolic links built by ICU for the common (‘uc’) library, version 5.4.3, for Linux
| || ||Required for link: Applications compiled with ‘ -licuuc’ will follow this symlink.|
| || ||Required for runtime: This name is what applications actually link against.|
| ||Actual library||Required for runtime and link. Contains the name |
Note: This discussion gives Linux as an example, but it is typical for most platforms, of which AIX and 390 (zOS) are exceptions.
An application compiled with ‘-licuuc’ will follow the symlink from
libicuuc.so.54.3, and will actually read the file
libicuuc.so.54.3. (fully qualified). This library file has an embedded name (SONAME) of
libicuuc.so.54, that is, with only the major and minor number. The linker will write this name into the client application, because Binary compatibility is for versions that share the same major+minor number.
If ICU version 5.4.7 is subsequently installed, the following files may be updated.
| || ||Required for link: Newly linked applications will follow this link, which should not cause any functional difference at link time.|
|libicuuc.so.54 ||Required for runtime: Because it now links to version .7, existing applications linked to version 5.4.3 will follow this link and use the 5.4.7 code.|
| ||Actual library||Required for runtime and link. Contains the name |
If ICU version 5.6.3 or 3.2.9 were installed, they would not affect already-linked applications, because the major+minor numbers are different - 56 and 32, respectively, as opposed to 54. They would, however, replace the link
libicuuc.so, which controls which version of ICU newly-linked applications use.
In summary, what files should an application distribute in order to include a functional runtime copy of ICU 5.4.3? The above application should distribute
libicuuc.so.54.3 and the symbolic link
libicuuc.so.54. (If symbolic links pose difficulty,
libicuuc.so.54.3 may be renamed to
libicuuc.so.54, and only
libicuuc.so.54 distributed. This is less informative, but functional.)
The –with-library-suffix option may be used with runConfigureICU or configure, to distinguish on disk specially modified versions of ICU. For example, the option –with-library-suffix=myapp will produce libraries with names such as libicuucmyapp.so.54.3, thus preventing another ICU user from using myapp’s custom ICU libraries.
While two or more versions of ICU may be linked into the same application as long as the major and minor numbers are different, changing the library suffix is not sufficient to allow the same version of ICU to be linked. In other words, linking ICU 5.4.3, 5.6.3, and 3.2.9 together is allowed, but 5.4.3 and 5.4.7 may not be linked together, nor may 5.4.3 and 5.4.3-myapp be linked together.
Assuming ICU version 5.4.3, Windows library names will follow this pattern:
||Release Link-time library. Needed for development. Contains |
| ||Release runtime library. Needed for runtime.|
||Debug link-time library (The |
||Debug runtime library.|
Debug applications must be linked with debug libraries, and release applications with release libraries.
When a new version of ICU is installed, the .lib files will be replaced so as to keep new compiles in sync with the newly installed header files, and the latest DLL. As well, if the new ICU version has the same major+minor version (such as 5.4.7), then DLLs will be replaced, as they are binary compatible. However, if an ICU with a different major+minor version is installed, such as 5.5, then new DLLs will be copied with names such as ‘icuuc55.dll’.
The services which are now known as ICU were written to provide operating system-level and application environment-level services. Several operating systems include ICU as a standard or optional package. See ICU Binary Compatibility for more details.