[PATCH] introduce upath repositories (issue793)

[Adrian Buehlmann]

On 18.06.2008 14:49, Paul Moore wrote:
> On 18/06/2008, Adrian Buehlmann <adrian at cadifra.com> wrote:
>> As mpm laid out, best disk performance is achieved if the files in the
>> store and in the working dir are in the same directory order
>> (mpm: "Ideally, the store layout should roughly mirror the working dir layout").
>>
>> The current filename encoding (also upath as presented) uses "_x"
>> to encode "X". Wouldn't it be a bit more efficient to do "X" -> "x_"?
>>
>> What was the reasoning for chosing "X" -> "_x" and not "X" -> "x_"?
> 
> My guess (and it's nothing more than that) is that it's related to this:
> 
>>>> 'x' < 'X'
> False
>>>> 'x' < '_x'
> False
>>>> 'x' < 'x_'
> True
> 
> Paul.
> 

Sorry, my bad. It seems the current encoding *is* actually the best with
regards to ordering. At least an ls -l on FreeBSD in the working dir and inside
.hg/store/data of the hg repo lists the files in the same order.

Interesting thing is, dir command on Windows lists them in different order:

> dir
 Volume in drive W is Sys
 Volume Serial Number is 8017-C29E

 Directory of W:\hgrepo

18.06.2008  19:51    <DIR>          .
18.06.2008  19:51    <DIR>          ..
18.06.2008  19:51    <DIR>          .hg
18.06.2008  19:51               326 .hgignore
18.06.2008  19:51             1'056 .hgsigs
18.06.2008  19:51               921 .hgtags
18.06.2008  19:51    <DIR>          contrib
18.06.2008  19:51             1'663 CONTRIBUTORS
18.06.2008  19:51            17'992 COPYING
18.06.2008  19:51    <DIR>          doc
18.06.2008  19:51               537 hg
18.06.2008  19:51             1'251 hgeditor
18.06.2008  19:51    <DIR>          hgext
18.06.2008  19:51               886 hgweb.cgi
18.06.2008  19:51             1'506 hgwebdir.cgi
18.06.2008  19:51             2'277 Makefile
18.06.2008  19:51    <DIR>          mercurial
18.06.2008  19:51               310 README
18.06.2008  19:51             4'631 setup.py
18.06.2008  19:51    <DIR>          templates
18.06.2008  19:51    <DIR>          tests
              12 File(s)         33'356 bytes
               9 Dir(s)  31'105'040'384 bytes free

> cd .hg\store\data

> dir
 Volume in drive W is Sys
 Volume Serial Number is 8017-C29E

 Directory of W:\hgrepo\.hg\store\data

18.06.2008  19:51    <DIR>          .
18.06.2008  19:51    <DIR>          ..
18.06.2008  19:50             4'193 .hgignore.i
18.06.2008  19:50             1'653 .hgsigs.i
18.06.2008  19:50             2'207 .hgtags.i
18.06.2008  19:50               977 comparison.txt.i
18.06.2008  19:50    <DIR>          contrib
18.06.2008  19:50               529 convert-repo.i
18.06.2008  19:50    <DIR>          doc
18.06.2008  19:50            27'362 hg.i
18.06.2008  19:50             5'988 hgeditor.i
18.06.2008  19:50    <DIR>          hgext
18.06.2008  19:50            11'219 hgmerge.i
18.06.2008  19:50             3'033 hgweb.cgi.i
18.06.2008  19:50             6'746 hgweb.py.i
18.06.2008  19:50             3'168 hgwebdir.cgi.i
18.06.2008  19:50    <DIR>          mercurial
18.06.2008  19:50             5'018 notes.txt.i
18.06.2008  19:50             5'244 purge.py.i
18.06.2008  19:50               321 rewrite-log.i
18.06.2008  19:50            13'997 setup.py.i
18.06.2008  19:50    <DIR>          templates
18.06.2008  19:50    <DIR>          tests
18.06.2008  19:51               110 tkmerge.i
18.06.2008  19:50             2'993 _c_o_n_t_r_i_b_u_t_o_r_s.i
18.06.2008  19:50             6'901 _c_o_p_y_i_n_g.i
18.06.2008  19:50             6'901 _c_o_p_y_i_n_g._g_p_l.i
18.06.2008  19:50             3'608 _makefile.i
18.06.2008  19:50             3'688 _m_a_n_i_f_e_s_t.in.i
18.06.2008  19:50               418 _p_k_g-_i_n_f_o.i
18.06.2008  19:50            11'330 _r_e_a_d_m_e.i
18.06.2008  19:50             5'620 _t_o_d_o.i
              24 File(s)        133'224 bytes
               8 Dir(s)  31'105'044'480 bytes free

But since the whole block of the uppercase entries is most likely kept
together it probably doesn't matter much where that block is moved to.

It's most likely the best thing to have all the uppercase entries
together. And that's what we have with the current encoding.

So my question was stupid and useless. Sorry for the noise.