On Mon, Dec 5, 2011 at 10:44 PM, Karl Ha= mmar <karl AT aspodata DOT se> wrote:

Luigi:

...
> In attachment an improved version of the script.

...

That version works fine.

It seems a little strange, though, to use both ftw and opendir.
Let me to explain why this decision. In my point of view we have = two targets:

a) build a list with the sub-folders, and
b) skip su= b-folder with no files or files other than .sym

So the ftw system is very fast to build the tree of sub-folders, but it= scans all the tree in depth. I haven't found any option to stop it at = the first level of the tree. This is the real reason on the why i use readd= ir in the (is-there-any-symbols-in?) procedure. readdir,instead, stops at t= he first level of the tree and the procedure exit immediately when found a = sym file.
So the code in (build-symbols-list) does a searching for only the subfolder= s containing sym files and exiting as fast as possible.
I thought the overhead of this mechanism is lower than an hash calculation= for each sym file. But I was wrong !
So to verify this i make a test as follow:

i have downloaded all ged= asymbols site in the folder gedasymbol and copy it 4 times so i have:
ge= dasymbol
gedasymbol2
gedasymbol3
gedasymbol4
gedasymbol5

main folders. Each folder has (at now):
716 folder (x5 3580)
4997 files of this 1754 are symbols (x5 24985 8770)=
=A0
in gafrc i put the following lines:
(load (build-path geda-rc= home-path "local-symbols-library.scm")) <-- opendir version ;;;(load (build-path geda-rchome-path "peter.scm")) <-- Peter= =A0 version
;;;(load (build-path geda-rchome-path "karl.scm"))= =A0 <-- Karl=A0 version
;;;(load (build-path geda-rchome-path "l= ocal-symbols-library-hash-version.scm")) <-- we'll see later
(component-library-add-tree "Own" "/home/ultrabit/Progra= mmingWorkspace/gitrepos/sviluppo-100/gitosis-admin/Footprints/gEda-sym/pers= onal/")
(component-library-add-tree "gEDA" "/home/ultrabit/Programmi= ngWorkspace/electronics/geda-suite/gedasymbols/www/user")
(componen= t-library-add-tree "gEDA1" "/home/ultrabit/ProgrammingWorksp= ace/electronics/geda-suite/gedasymbols2/www/user")
(component-library-add-tree "gEDA2" "/home/ultrabit/Programm= ingWorkspace/electronics/geda-suite/gedasymbols3/www/user")
(compon= ent-library-add-tree "gEDA3" "/home/ultrabit/ProgrammingWork= space/electronics/geda-suite/gedasymbols4/www/user")
(component-library-add-tree "gEDA4" "/home/ultrabit/Programm= ingWorkspace/electronics/geda-suite/gedasymbols5/www/user")

at = the prompt shell i run:

ultrabit AT hurricane:~/.gEDA$ time gschem -c &= quot;(gschem-exit)"=A0=A0=A0=A0 <--- loading gschem and exit at sta= rtup
Invalid path [./sym] passed to component-library
Invalid path [./footpri= nt] passed to component-library
Invalid path [./footprint] passed to com= ponent-library-search

real=A0=A0=A0 0m1.427s=A0=A0
user=A0=A0=A0 0m1.100s
sys=A0=A0=A0 0m0.248s

(total gschem loadin= g time (intel core 2 @ 3Ghz and 8 Gb of ram (not all free))

Step 2 = second test some conditions but Peter version:
;;;(load (build-path geda= -rchome-path "local-symbols-library.scm")) <-- opendir version=
(load (build-path geda-rchome-path "peter.scm")) <-- Peter=A0 = version
;;;(load (build-path geda-rchome-path "karl.scm"))=A0 <-- Karl= =A0 version
;;;(load (build-path geda-rchome-path "local-symbols-library-hash-vers= ion.scm")) <-- we'll see later
=A0
ultrabit AT hurricane:~/.= gEDA$ time gschem -c "(gschem-exit)"
Invalid path [./sym] pass= ed to component-library
Invalid path [./footprint] passed to component-library
Invalid path [./f= ootprint] passed to component-library-search

real=A0=A0=A0 0m5.538s<= br>user=A0=A0=A0 0m3.600s
sys=A0=A0=A0 0m1.840s

Step 3 same condi= tions but with Karl version:
ultrabit AT hurricane:~/.gEDA$ time gschem -c "(gschem-exit)"
Inv= alid path [./sym] passed to component-library
Invalid path [./footprint]= passed to component-library
Invalid path [./footprint] passed to compon= ent-library-search

real=A0=A0=A0 0m1.436s
user=A0=A0=A0 0m1.172s
sys=A0=A0=A0 0m0.16= 0s

So, i am surprised about results because i thought the Karl versi= on was like the Peter version, but this is not true.
I found the proble= m (many thanks to the "two column" emacs mode) was the use of (ha= sx-set!) vs (hashq-set!) and storing symbols instead of strings .
In conclusion i made a very condensed guile code (one procedure) to do all = using only ftw:

Step 4:
;;;(load (build-path geda-rchome-path &qu= ot;local-symbols-library.scm")) <-- opendir version
;;;(load (build-path geda-rchome-path "peter.scm")) <-- Peter= =A0 version
;;;(load (build-path geda-rchome-path "karl.scm"))=A0 <-- Karl= =A0 version
(load (build-path geda-rchome-path "local-symbols-library.scm"))= =A0 <-- Optimized hash version

ultrabit AT hurricane:~/.gEDA$ time g= schem -c "(gschem-exit)"
Invalid path [./sym] passed to compon= ent-library
Invalid path [./footprint] passed to component-library
Invalid path [./f= ootprint] passed to component-library-search

real=A0=A0=A0 0m1.229s<= br>user=A0=A0=A0 0m0.980s
sys=A0=A0=A0 0m0.152s

The better i can = do :-D

If we are going to use opendir, we could just as well do the
traversal ourself and dump ftw.

I see two alternatives:

a) I'm satisfied with my version, and stop here

b) we
=A0 =A0dump ftw
=A0 =A0do the traversal ourself
=A0 =A0clean up your script
=A0and proposes it for inclusion in geda

What do you think?

Karl, please look at code in at= tachment and say me if can be accepted as is. I think this can be proposed = for inclusion in geda.

=A0

Regards,
/Karl Hammar

-----------------------------------------------------------------------
Asp=F6 Data
Lilla Asp=F6 148
S-742 94 =D6sthammar
Sweden
+46 173 140 57