Manual browser: xstr(1)

XSTR(1) General Commands Manual XSTR(1)


xstrextract strings from C programs to implement shared strings


xstr [-cv] [-l array] [-] [file ...]


xstr maintains a file strings into which strings in component parts of a large program are hashed. These strings are replaced with references to this common area. This serves to implement shared constant strings, most useful if they are also read-only.

Available options:

xstr reads from the standard input.
xstr will extract the strings from the C source file or the standard input (-), replacing string references by expressions of the form (&xstr[number]) for some number. An appropriate declaration of xstr is prepended to the file. The resulting C text is placed in the file x.c, to then be compiled. The strings from this file are placed in the strings data base if they are not there already. Repeated strings and strings which are suffixes of existing strings do not cause changes to the data base.
-l array
Specify the named array in program references to abstracted strings. The default array name is xstr.
Be verbose.

After all components of a large program have been compiled, a file xs.c declaring the common xstr space can be created by a command of the form:

$ xstr

The file xs.c should then be compiled and loaded with the rest of the program. If possible, the array can be made read-only (shared) saving space and swap overhead.

xstr can also be used on a single file. The following command creates files x.c and xs.c as before, without using or affecting any strings file in the same directory:

$ xstr name

It may be useful to run xstr after the C preprocessor if any macro definitions yield strings or if there is conditional code which contains strings which may not, in fact, be needed. An appropriate command sequence for running xstr after the C preprocessor is:

$ cc -E name.c | xstr -c - 
$ cc -c x.c 
$ mv x.o name.o

xstr does not touch the file strings unless new items are added, thus make(1) can avoid remaking xs.o unless truly necessary.


Data base of strings
Massaged C source
C source for definition of array `xstr'
Temp file when `xstr name' doesn't touch strings




The xstr command appeared in 3.0BSD.


If a string is a suffix of another string in the data base, but the shorter string is seen first by xstr both strings will be placed in the data base, when just placing the longer one there will do.

xstr does not parse the file properly so it does not know not to process:

	char var[] = "const";

	char var[] = (&xstr[N]);

These must be changed manually into an appropriate initialization for the string, or use the following ugly hack.

Also, xstr cannot initialize structures and unions that contain strings. Those can be fixed by changing from:

	struct foo { 
		int i; 
		char buf[10]; 
	} = { 
		1, "foo" 

	struct foo { 
		int i; 
		char buf[10]; 
	} = { 
		1, { 'f', 'o', 'o', '\0' } 

The real problem in both cases above is that the compiler knows the size of the literal constant so that it can perform the initialization required, but when xstr changes the literal string to a pointer reference, the size information is lost. It would require a real parser to do this right, so the obvious solution is to fix the program manually to compile, or even better rely on the compiler and the linker to merge strings appropriately.

Finally, xstr is not very useful these days because most of the string merging is done automatically by the compiler and the linker, provided that the strings are identical and read-only.

July 23, 2004 NetBSD 7.0