a little archeology in the repository

hi folks,

im writhing a script that pulls code out of a
set of old snapshot directories (and any RCS/*,v files
within them) to try to organize the chaos that
is my hard-drive.

once the revs are in svn, I do a `svn diff -r$A:$B $file`
inside a double loop, and count the ^[-+] lines of each
to produce the following;

looking at ebox.pl
  revs: 13 12 11 10 9 8 7 6 5 4
3 1
    --- --- --- --- --- --- --- --- --- --- ---
--- ---
    13: 0 47 49 51 66 107 138 140 156 203
215 0
    12: 0 2 4 19 62 95 97 113 162
178 47
    11: 0 2 17 60 93 95 111 160
176 49
    10: 0 15 58 91 93 109 158
174 51
     9: 0 43 78 80 96 149
163 66
     8: 0 45 47 63 122
136 107
     7: 0 2 18 109
123 138
     6: 0 18 109
123 140
     5: 0 105
119 156
     4: 0
26 203
     3:
0 215
     
1: 0

Heres a few things that can be seen here (if you look hard):

1. can estimate thrashing...

-r13:12 shows 47 changes,
-r12:11 shows 2.
-r13:11 shows 49,

this indicates that theres no overlap between the changes, otherwize
-r13:11 would show <49

same pattern is evident with 12,11,10

on the other hand,
 -r13:11=49,
 -r12:10=4,
 -r13:10=51,
shows that 2 lines were thrashed - changed in both revisions.

2. 1st rev was latest.

last column, comparing rev1 to all others, has changes in reverse.
this is cuz rev1 was taken from the plain-file, others were taken from
the RCS/ebox.pl,v file, from oldest to youngest. The youngest
matches rev1; I added it into RCS if it didnt already match the latest
rcs-rev.

3. monotonic increase in changes

in other columns, numbers increase as you go up the column,
indicating that changes accumulated, nothing was reverted.

This is partly an artifact - as the script iterates thru each directory,
it checksums each file (and each rev pulled from RCS) to see if it
was already SVNd, thus duplicates are avoided. If this wasnt the case,
then the matrix would have another diagonal of 0s, with same numbers
repeating..

RCS/ebox.pl,v --> ebox.pl
revision 1.10
done
already seen this: echobox.19981019/ebox.pl sum 144644385417483

4. version to version changes are on the diagonal just above the
identity one.

ie -r$REV+1:$REV

So, the Questions.

1. anybody see a better way to glean an understanding of history ?

2. will SVN ever support a re-upgrade ala my checksum check ?
(post 1.0 if ever, probably)

3. is there any value to having this script in examples ?
Received on Fri Aug 22 09:05:31 2003