Re: About version sorting of `ls'.

[Jim Meyering]

Thank you for the report and patch.
Since strverscmp.c is part of the GNU C library,
I'll pass this on to them.

ISONO Tomoatsu <address@hidden> wrote:
| Dear Maintainers,
|
|   In fileutils-4.0.41, there is a conflict between the description of
| version sorting and a result of it.  The conflict concerns fractional
| cases.
|
| # SunOS smaux 5.7 Generic_106541-05 sun4u sparc
|
|     A result of `ls -1v'.  The description in fileutils.texi.
|      abc-1.007.tgz          abc-1.007.tgz
|      abc-1.012b.tgz         abc-1.01a.tgz
|      abc-1.01a.tgz          abc-1.012b.tgz
|
|   A resolution is given below, which makes `ls' output in the
| described way.  I hope you will adopt this kind of change.
|
|                                        Yours sincerely,
|                                          ISONO Tomoatsu
| -------------------------------------------------------------------
|  ISONO Tomoatsu <address@hidden>
| -------------------------------------------------------------------
|
|
| diff -rc fileutils-4.0.41/lib/strverscmp.c 
fileutils-4.0.41+modified-20010305/lib/strverscmp.c
| *** fileutils-4.0.41/lib/strverscmp.c Thu Nov 16 18:08:53 2000
| --- fileutils-4.0.41+modified-20010305/lib/strverscmp.c       Mon Mar  5 
08:15:11 2001
| ***************
| *** 36,41 ****
| --- 36,43 ----
|   #define CMP    2
|   #define LEN    3
|   
| + /* Use DUM to fill spaces. */
| + #define DUM    0
|   
|   /* ISDIGIT differs from isdigit, as follows:
|      - Its arg may be any int or unsigned int; it need not be an unsigned 
char.
| ***************
| *** 59,64 ****
| --- 61,119 ----
|      equal to or greater than S2 (for more info, see the texinfo doc).
|   */
|   
| + /* MODIFIED VERSION:
| +    The following is the ordering of the original version.
| + 
| +    a.x
| +    a00.x
| +    a01.x
| +    a012.x
| +    a01a.x
| +    a0a.x
| +    aa.x
| +    abc-1.007.tgz
| +    abc-1.012b.tgz
| +    abc-1.01a.tgz
| + 
| +    But I think this is not so good, and suggest a new clear definition of
| +    order.  And, please note that the last three filenames are examples in
| +    `doc/fileutils.texi' and that the order of them is different from the
| +    description there.
| + 
| +    1. Treat each maximal successive sequence of digits /[0-9]+/ as a
| +       `character' (we refer it as DIGSEQ).
| + 
| +    2. The numerical value of a DIGSEQ is considered as a fractional one if
| +       it begins with `0'. But in fractional cases, there exists ambiguity
| +       due to last successive `0's in DIGSEQ. So, when the numerical values
| +       of them are equal, we regard the longer one as the larger one.
| + 
| +       (This is equivalent to consider as following: we define numerical
| +       values of `0' (resp. `1',... ,`9') as 1 (resp. 2,... , 10). Consider
| +       a DIGSEQ is an undecimal, or eleven-adic expression (both in integer
| +       case and in fractional case).
| + 
| +    3. The order between a DIGSEQ and a non-digit character CHAR is defined
| +       to be the same order between '0' and CHAR.
| +    
| +    4. The order between two DIGSEQs is defined to be the same order as
| +       numerical values of two DIGSEQs.
| + 
| +    5. Compare strings according to the order of `character's.
| + 
| +    Original        Modified
| +     a.x             a.x
| +     a00.x           a0a.x
| +     a01.x           a00.x
| +     a012.x          a01.x
| +     a01a.x          a01a.x
| +     a0a.x           a012.x
| +     aa.x            aa.x
| +     abc-1.007.tgz   abc-1.007.tgz
| +     abc-1.012b.tgz  abc-1.01a.tgz
| +     abc-1.01a.tgz   abc-1.012b.tgz
| + */
| + 
|   int
|   __strverscmp (const char *s1, const char *s2)
|   {
| ***************
| *** 73,89 ****
| --- 128,164 ----
|     static const unsigned int next_state[] =
|     {
|         /* state    x    d    0    - */
| +     /* the first zero digit means fractional state,
| +        the first non-zero digit means integral state. */
| +       /* S_N */  S_N, S_I, S_F, DUM, 
| +     /* the successive digit means integral in integral state. */
| +       /* S_I */  S_N, S_I, S_I, DUM,
| +     /* the successive digit means fractional in fractional state. */
| +       /* S_F */  S_N, S_F, S_F/*,DUM,*/
| +     /* No need to use S_Z state */
| + #if 0
|         /* S_N */  S_N, S_I, S_Z, S_N,
|         /* S_I */  S_N, S_I, S_I, S_I,
|         /* S_F */  S_N, S_F, S_F, S_F,
|         /* S_Z */  S_N, S_F, S_Z, S_Z
| + #endif
|     };
|   
|     static const int result_type[] =
|     {
| +     /* Note we never refer to 0/0 */
|         /* state   x/x  x/d  x/0  x/-  d/x  d/d  d/0  d/-
|                    0/x  0/d  0/0  0/-  -/x  -/d  -/0  -/- */
|   
| +       /* S_N */  CMP, CMP, CMP, DUM, CMP, LEN, CMP, DUM, /* int > frac */
| +                  CMP, CMP, DUM, DUM, DUM, DUM, DUM, DUM, 
| +       /* S_I */  CMP, -1,  -1,  DUM,  1,  LEN, LEN, DUM, 
| +                   1,  LEN, DUM, DUM, DUM, DUM, DUM, DUM, 
| +       /* S_F */  CMP, -1,  -1,  DUM,  1,  CMP, CMP, DUM, /* no need LEN cmp 
*/
| +                   1,  CMP/*,DUM,DUM, DUM, DUM, DUM, DUM, */
| +       /* S_Z *//*CMP, -1,  -1,  DUM,  1,  CMP, CMP, DUM, 
| +                   1,  CMP, CMP, DUM */
| + #if 0
|         /* S_N */  CMP, CMP, CMP, CMP, CMP, LEN, CMP, CMP,
|                    CMP, CMP, CMP, CMP, CMP, CMP, CMP, CMP,
|         /* S_I */  CMP, -1,  -1,  CMP,  1,  LEN, LEN, CMP,
| ***************
| *** 92,97 ****
| --- 167,173 ----
|                    CMP, CMP, CMP, CMP, CMP, CMP, CMP, CMP,
|         /* S_Z */  CMP,  1,   1,  CMP, -1,  CMP, CMP, CMP,
|                    -1,  CMP, CMP, CMP
| + #endif
|     };
|   
|     if (p1 == p2)
| 
| _______________________________________________
| Bug-fileutils mailing list
| address@hidden
| http://mail.gnu.org/mailman/listinfo/bug-fileutils