libfsd-devel Mailing List for Finite State Devices Library
Status: Planning
Brought to you by:
vanio
Menu
▾
▴
libfsd-devel — Development issues
You can subscribe to this list here.
| 2005 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(19) |
Oct
(16) |
Nov
(1) |
Dec
|
|---|
|
From: Ivan P. <va...@ne...> - 2005-11-03 07:55:23
|
Snoshti napravih determinizacija na NFA (koeto e vsushtnost konstruktora na
DFA ot NFA). Pichove, trjabva da vi kazha che otdavna ne sum pisal neshto
deto da me nakara da mislja tolkova. Hem go znam algorituma do poslednija
detail i vse pak na praktika mi se oteli vola dokato go napravja. Krajnija
rezultat e baja izchisten, makar i da turpi kritiki otkum efektivnost.
Oshte ne sum commitnal (beshe 2.30 i sled purvite njakolko determiniriani
avtomata prosto pripadnah v legloto) no vi prashtam primer - "x(a*|b*)y"
suotvetno NFA i DFA.
Osven tova neshto interesno e worst case-a na algorituma za determinizacija --
deto vodi do exponencialen vzriv na sustojanijata. Primera e napravo velik!
Reguljarnijat izraz ".*1[01]{12}" generira NFA s 29 sustojanija i DFA s
8193! :) Intuitivno tozi izraz kazva "vsichki dumi, koito imat edinica 12
pozicii predi kraja". I ponezhe avtomata ne mozhe da gleda napred, toj
trjabva da pomni kakvo e stanalo predi 12 pozicii. I ponezhe na nego pametta
sa mu sustojanijata, toj trjabva da ima put v grafa si za vseki vuzmozhen
razvoj, demek vsichki vectori ot nuli e edinici s razmernost 12. Seshtate
se.. Za shtastie takiva reguljarni izrazi mnogo rjadko se sreshtat i v 99% ot
sluchaite vsichko e mnogo burzo i dovolno.
|
|
From: Ivan P. <va...@ne...> - 2005-10-28 07:06:51
|
On Tuesday 25 October 2005 10:54, Georgi Mungov wrote:
> Ami moje bi ako imame serialize/deserialize methods i testa da
> predstavliava generirane na avtomat ot regex, serialize i compare s neshto
> koeto e provereno predi tova che e viarno. Loshoto e che pak zavisi ot
> strukturata na avtomata
>
> Drugoto neshto koeto zabeliazah e che polimorfizma gfa<-nfa<-dfa e malko
> kuc poneje suotvetnite structuri niamat ukazatel kum metodite si. Suotvetno
> v NFA i DFA ima mnogo metodi koito prosto vikat suotvetnia na roditelia.
> Neznam dali niakoga shte se naloji da se polzva polimorfizum i ako da kolko
> po elegantno moje da se napravi na C bez external OO framework i bez da se
> pishe mnogo kod. Varianta
>
> dfa->method(dfa, ....)
>
> pak ne e koi znae kolko eleganten
>
abe ako ne drugo sum napravil metodite deto vikat metodite na roditelja da sa
inline, ta obshto vzeto ne bi trjabvalo da ima overhead.. mda. obektno na C
vinagi si e bilo kutzo na vid...
> Vanio ti update-na li si cv-to:)
mi chakam i Pesho da kazhe kak e tam :))
>
> On Thursday 13 October 2005 15:45, Ivan Peikov wrote:
> > ima novosti v CVS-a - naj-veche bugfixes. dobavih dfa.{c,h} s basic
> > interface-a (kuhi metodi) i v momenta se opitvam da napravija
> > konstruirane ot NFA (demek determinizacija).
> >
> > kato go svursha tova mislja da pospra s kodiraneto, zashtoto mnogo se
> > natrupaha neshtata i stava vse po-nevuzmozhno da se napishat autotestove.
> >
> > ta kato stana duma za autotests, mnogo mislih kak mozhe da se definirat.
> > znachi imame njakakvi grafi, obache reda na transitionite ili
> > nomeracijata na sustojanijata ne e ot znachenie. znachi promerno mozhem
> > da zadadem grafa na avtomata po njakakuv nachin (XML?) obache sled tova
> > ne trjabva da durzhim na naredbi. toest opirame do izomorfizum na grafi,
> > puk tova pochti sigurno e NP-complete (maj beshe otvoren problema). ne
> > che shte e grizha za malki grafcheta, no vse pak.. tova mi se vizhda
> > edinstvenija kaduren variant kojto da proverjava strukturata na
> > generiranite avtomati. drug variant za testvane, no ne tolkova
> > izcherpatelen e da zadavame reguljaren izraz i njakakuv broj accepted i
> > non-accapted putishta. toest za reguljarnija izraz a+b mozhem da zadadem
> > che generiranija avtomat trjabva da razpoznava aaaaaaaab i ab, no trjabva
> > da othvyrli a, abb, bb, x i oshte mnogo neshta.
> > tova ima predimstvoto che njama da ni vurzhe strukturata na generiranite
> > avtomati, t.e. ako utre smenim algorituma za konstruirane njama da
> > trjabva da prepravjame testovete, no puk ne e tolkova sigurno. ama puk
> > kato se zamislja mnogo lesno shte se dobavjat test case-ove.. ne znam -
> > kakvo mislite?
> >
> > za reguljarnite izrazi kato che li naj-dobre e da se opisva
> > parse-durvoto. tam taka ili inache naredbata na decata ima znachenie...
> >
> > On Friday 07 October 2005 09:38, Ivan Peikov wrote:
> > > Commitnal sum poslednata mi versija ot vkushti.
> > > Edinstvenata po-goljama promjana e che v transitionite (nfa_trans_t)
> > > veche ima ne simvoli (nfa_alpha_t), range-ove (nfa_range_t). Suotvetno
> > > konstrukcijata e leko promenena (i sushto taka zavurshena -- predi
> > > njamashe support za character lists vuv REX parse durvetata).
> > > Taka che sega vseki reguljaren izraz kojto se poddurzha ot REX parsera
> > > si ima ekvivalenten avtomat.
> > >
> > > Attachvam kakvo se generira za reguljarnija izraz
> > > "[[:alpha:]][[:alnum:]_]*". Rezultatite mislja sa neprilichno dobri.
> > > Parsvaneto (na NC-mashinata mi) na tozi reguljaren izraz izjazhda
> > > ~20usec, a NFA konstrukcijata po tozi REX ~60usec. Tova oznachava che
> > > ima megdan za determinirane i pak shte sme v obozrimi granici. Pametta
> > > ne sum ja gledal, no kakto pisah predi, garantirano e che e linejna po
> > > reguljarnija izraz ako njama iteracija, i e kvadratichna, kogato ima.
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by:
> > Power Architecture Resource Center: Free content, downloads, discussions,
> > and more. http://solutions.newsforge.com/ibmarch.tmpl
> > _______________________________________________
> > Libfsd-devel mailing list
> > Lib...@li...
> > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
>
> "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following
> annotations on 10/25/05 01:04:44
> ---------------------------------------------------------------------------
>--- This e-mail, including attachments, may include confidential and/or
> proprietary information, and may be used only by the person or entity to
> which it is addressed. If the reader of this e-mail is not the intended
> recipient or his or her authorized agent, the reader is hereby notified
> that any dissemination, distribution or copying of this e-mail is
> prohibited. If you have received this e-mail in error, please notify the
> sender by replying to this message and delete this e-mail immediately.
> ===========================================================================
>===
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by the JBoss Inc.
> Get Certified Today * Register for a JBoss Training Course
> Free Certification Exam for All Training Attendees Through End of 2005
> Visit http://www.jboss.com/services/certification for more information
> _______________________________________________
> Libfsd-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libfsd-devel
|
|
From: Georgi M. <geo...@tu...> - 2005-10-25 07:56:07
|
Ami moje bi ako imame serialize/deserialize methods i testa da predstavliava
generirane na avtomat ot regex, serialize i compare s neshto koeto e provereno
predi tova che e viarno. Loshoto e che pak zavisi ot strukturata na avtomata
Drugoto neshto koeto zabeliazah e che polimorfizma gfa<-nfa<-dfa e malko kuc
poneje suotvetnite structuri niamat ukazatel kum metodite si. Suotvetno v NFA
i DFA ima mnogo metodi koito prosto vikat suotvetnia na roditelia. Neznam
dali niakoga shte se naloji da se polzva polimorfizum i ako da kolko po
elegantno moje da se napravi na C bez external OO framework i bez da se pishe
mnogo kod. Varianta
dfa->method(dfa, ....)
pak ne e koi znae kolko eleganten
Vanio ti update-na li si cv-to:)
On Thursday 13 October 2005 15:45, Ivan Peikov wrote:
> ima novosti v CVS-a - naj-veche bugfixes. dobavih dfa.{c,h} s basic
> interface-a (kuhi metodi) i v momenta se opitvam da napravija konstruirane
> ot NFA (demek determinizacija).
>
> kato go svursha tova mislja da pospra s kodiraneto, zashtoto mnogo se
> natrupaha neshtata i stava vse po-nevuzmozhno da se napishat autotestove.
>
> ta kato stana duma za autotests, mnogo mislih kak mozhe da se definirat.
> znachi imame njakakvi grafi, obache reda na transitionite ili nomeracijata
> na sustojanijata ne e ot znachenie. znachi promerno mozhem da zadadem grafa
> na avtomata po njakakuv nachin (XML?) obache sled tova ne trjabva da
> durzhim na naredbi. toest opirame do izomorfizum na grafi, puk tova pochti
> sigurno e NP-complete (maj beshe otvoren problema). ne che shte e grizha za
> malki grafcheta, no vse pak.. tova mi se vizhda edinstvenija kaduren
> variant kojto da proverjava strukturata na generiranite avtomati. drug
> variant za testvane, no ne tolkova izcherpatelen e da zadavame reguljaren
> izraz i njakakuv broj accepted i non-accapted putishta. toest za
> reguljarnija izraz a+b mozhem da zadadem che generiranija avtomat trjabva
> da razpoznava aaaaaaaab i ab, no trjabva da othvyrli a, abb, bb, x i oshte
> mnogo neshta.
> tova ima predimstvoto che njama da ni vurzhe strukturata na generiranite
> avtomati, t.e. ako utre smenim algorituma za konstruirane njama da trjabva
> da prepravjame testovete, no puk ne e tolkova sigurno. ama puk kato se
> zamislja mnogo lesno shte se dobavjat test case-ove.. ne znam - kakvo
> mislite?
>
> za reguljarnite izrazi kato che li naj-dobre e da se opisva parse-durvoto.
> tam taka ili inache naredbata na decata ima znachenie...
>
> On Friday 07 October 2005 09:38, Ivan Peikov wrote:
> > Commitnal sum poslednata mi versija ot vkushti.
> > Edinstvenata po-goljama promjana e che v transitionite (nfa_trans_t)
> > veche ima ne simvoli (nfa_alpha_t), range-ove (nfa_range_t). Suotvetno
> > konstrukcijata e leko promenena (i sushto taka zavurshena -- predi
> > njamashe support za character lists vuv REX parse durvetata).
> > Taka che sega vseki reguljaren izraz kojto se poddurzha ot REX parsera si
> > ima ekvivalenten avtomat.
> >
> > Attachvam kakvo se generira za reguljarnija izraz
> > "[[:alpha:]][[:alnum:]_]*". Rezultatite mislja sa neprilichno dobri.
> > Parsvaneto (na NC-mashinata mi) na tozi reguljaren izraz izjazhda
> > ~20usec, a NFA konstrukcijata po tozi REX ~60usec. Tova oznachava che ima
> > megdan za determinirane i pak shte sme v obozrimi granici. Pametta ne sum
> > ja gledal, no kakto pisah predi, garantirano e che e linejna po
> > reguljarnija izraz ako njama iteracija, i e kvadratichna, kogato ima.
>
> -------------------------------------------------------
> This SF.Net email is sponsored by:
> Power Architecture Resource Center: Free content, downloads, discussions,
> and more. http://solutions.newsforge.com/ibmarch.tmpl
> _______________________________________________
> Libfsd-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libfsd-devel
"Tumbleweed E-mail Firewall <tumbleweed.com>" made the following
annotations on 10/25/05 01:04:44
------------------------------------------------------------------------------
This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately.
==============================================================================
|
|
From: Ivan P. <va...@ne...> - 2005-10-13 12:45:36
|
ima novosti v CVS-a - naj-veche bugfixes. dobavih dfa.{c,h} s basic
interface-a (kuhi metodi) i v momenta se opitvam da napravija konstruirane ot
NFA (demek determinizacija).
kato go svursha tova mislja da pospra s kodiraneto, zashtoto mnogo se
natrupaha neshtata i stava vse po-nevuzmozhno da se napishat autotestove.
ta kato stana duma za autotests, mnogo mislih kak mozhe da se definirat.
znachi imame njakakvi grafi, obache reda na transitionite ili nomeracijata na
sustojanijata ne e ot znachenie. znachi promerno mozhem da zadadem grafa na
avtomata po njakakuv nachin (XML?) obache sled tova ne trjabva da durzhim na
naredbi. toest opirame do izomorfizum na grafi, puk tova pochti sigurno e
NP-complete (maj beshe otvoren problema). ne che shte e grizha za malki
grafcheta, no vse pak.. tova mi se vizhda edinstvenija kaduren variant kojto
da proverjava strukturata na generiranite avtomati. drug variant za testvane,
no ne tolkova izcherpatelen e da zadavame reguljaren izraz i njakakuv broj
accepted i non-accapted putishta. toest za reguljarnija izraz a+b mozhem da
zadadem che generiranija avtomat trjabva da razpoznava aaaaaaaab i ab, no
trjabva da othvyrli a, abb, bb, x i oshte mnogo neshta.
tova ima predimstvoto che njama da ni vurzhe strukturata na generiranite
avtomati, t.e. ako utre smenim algorituma za konstruirane njama da trjabva da
prepravjame testovete, no puk ne e tolkova sigurno. ama puk kato se zamislja
mnogo lesno shte se dobavjat test case-ove.. ne znam - kakvo mislite?
za reguljarnite izrazi kato che li naj-dobre e da se opisva parse-durvoto. tam
taka ili inache naredbata na decata ima znachenie...
On Friday 07 October 2005 09:38, Ivan Peikov wrote:
> Commitnal sum poslednata mi versija ot vkushti.
> Edinstvenata po-goljama promjana e che v transitionite (nfa_trans_t) veche
> ima ne simvoli (nfa_alpha_t), range-ove (nfa_range_t). Suotvetno
> konstrukcijata e leko promenena (i sushto taka zavurshena -- predi njamashe
> support za character lists vuv REX parse durvetata).
> Taka che sega vseki reguljaren izraz kojto se poddurzha ot REX parsera si
> ima ekvivalenten avtomat.
>
> Attachvam kakvo se generira za reguljarnija izraz
> "[[:alpha:]][[:alnum:]_]*". Rezultatite mislja sa neprilichno dobri.
> Parsvaneto (na NC-mashinata mi) na tozi reguljaren izraz izjazhda ~20usec,
> a NFA konstrukcijata po tozi REX ~60usec. Tova oznachava che ima megdan za
> determinirane i pak shte sme v obozrimi granici. Pametta ne sum ja gledal,
> no kakto pisah predi, garantirano e che e linejna po reguljarnija izraz ako
> njama iteracija, i e kvadratichna, kogato ima.
|
|
From: Ivan P. <va...@ne...> - 2005-10-07 06:39:17
|
Commitnal sum poslednata mi versija ot vkushti. Edinstvenata po-goljama promjana e che v transitionite (nfa_trans_t) veche ima ne simvoli (nfa_alpha_t), range-ove (nfa_range_t). Suotvetno konstrukcijata e leko promenena (i sushto taka zavurshena -- predi njamashe support za character lists vuv REX parse durvetata). Taka che sega vseki reguljaren izraz kojto se poddurzha ot REX parsera si ima ekvivalenten avtomat. Attachvam kakvo se generira za reguljarnija izraz "[[:alpha:]][[:alnum:]_]*". Rezultatite mislja sa neprilichno dobri. Parsvaneto (na NC-mashinata mi) na tozi reguljaren izraz izjazhda ~20usec, a NFA konstrukcijata po tozi REX ~60usec. Tova oznachava che ima megdan za determinirane i pak shte sme v obozrimi granici. Pametta ne sum ja gledal, no kakto pisah predi, garantirano e che e linejna po reguljarnija izraz ako njama iteracija, i e kvadratichna, kogato ima. |
|
From: Ivan P. <va...@ne...> - 2005-10-06 13:31:17
|
vutre si On Thursday 06 October 2005 15:58, Georgi Mungov wrote: > aide add-ni georgigm > > On Thursday 06 October 2005 11:13, Ivan Peikov wrote: > > ok, pichove, the fun starts here :) > > slozhih distroto koeto vi pratihp posledno v cvs-a na sourceforge.net. > > tova ne e poslednata versija, kojato imam vkushti (tja bachka s > > range-ovete za koito govorihme vchera -- stana mnoo jako) zashtoto FAT-a > > na mp3-playera mi reshi da mi pogodi nomer tazi sutrin. > > > > taka ili inache, za da mozhete da checkout-vate i da mazhete po cvs-a > > trjabva da si napravite user v SF.net i da mi go kazhete, posle az vi > > add-vam kum proekta i sled tova sus slednija magicheski red > > > > ==== > > cvs -z3 -d:ext:YO...@cv...:/cvsroot/libfsd checkout -d libfsd > > src ==== > > > > imate sandbox. > > > > > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: > > Power Architecture Resource Center: Free content, downloads, discussions, > > and more. http://solutions.newsforge.com/ibmarch.tmpl > > _______________________________________________ > > Libfsd-devel mailing list > > Lib...@li... > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > annotations on 10/06/05 06:07:49 > --------------------------------------------------------------------------- >--- This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified > that any dissemination, distribution or copying of this e-mail is > prohibited. If you have received this e-mail in error, please notify the > sender by replying to this message and delete this e-mail immediately. > =========================================================================== >=== > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel |
|
From: Georgi M. <geo...@tu...> - 2005-10-06 12:59:40
|
aide add-ni georgigm On Thursday 06 October 2005 11:13, Ivan Peikov wrote: > ok, pichove, the fun starts here :) > slozhih distroto koeto vi pratihp posledno v cvs-a na sourceforge.net. tova > ne e poslednata versija, kojato imam vkushti (tja bachka s range-ovete za > koito govorihme vchera -- stana mnoo jako) zashtoto FAT-a na mp3-playera mi > reshi da mi pogodi nomer tazi sutrin. > > taka ili inache, za da mozhete da checkout-vate i da mazhete po cvs-a > trjabva da si napravite user v SF.net i da mi go kazhete, posle az vi > add-vam kum proekta i sled tova sus slednija magicheski red > > ==== > cvs -z3 -d:ext:YO...@cv...:/cvsroot/libfsd checkout -d libfsd > src ==== > > imate sandbox. > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 10/06/05 06:07:49 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-10-06 08:13:47
|
ok, pichove, the fun starts here :) slozhih distroto koeto vi pratihp posledno v cvs-a na sourceforge.net. tova ne e poslednata versija, kojato imam vkushti (tja bachka s range-ovete za koito govorihme vchera -- stana mnoo jako) zashtoto FAT-a na mp3-playera mi reshi da mi pogodi nomer tazi sutrin. taka ili inache, za da mozhete da checkout-vate i da mazhete po cvs-a trjabva da si napravite user v SF.net i da mi go kazhete, posle az vi add-vam kum proekta i sled tova sus slednija magicheski red ==== cvs -z3 -d:ext:YO...@cv...:/cvsroot/libfsd checkout -d libfsd src ==== imate sandbox. |
|
From: Ivan P. <va...@ne...> - 2005-10-05 08:05:25
|
On Wednesday 05 October 2005 10:33, Georgi Mungov wrote: > Moje bi po skoro e list ot ranges zashtoto [0-9A-Za-z] sa realno 3 range-a > No togava mai shte stane mnogo tromavo. mhm, prav si - realno e list ot ranges, no ako slozhim list ot ranges v transition-a, naj-malkoto ni trjabva oshte edno nivo na vlozhenost, oshte edin pool za ranges i za tova mi se stori dobur kompromisa da ima samo po edin range v transition. pri tekushtata implementacija harchim mjasto kolkoto 6 pointera na transition i mislja che mozhem da preglutnem tova razhishtenie ako zhivota ni shte se oprosti > > Vidiah che si fixiral azbukata na ASCII. Mojem li da predstavim togava > range-a (ili ranges) sus bitset ne mi se iska da fixirame na ASCII, a predstavjane s bitset ni vruzva dosta seriozno da rabotim s 1-byte simvol (za Unicode bitseta bi bil 16K...) purvonachalnata ideja da typedef-na nfa_alpha_t beshe eventualna vuzmozhnost utre da go napravim wchar_t, i tazi vuzmozhnost zavisi edinstveno ot lipsata na konstrukcii, eksponencialni po sizeof(nfa_alpha_t) ne e kazano che sega trjabva da pravim unicode versija, no e hubavo pone da njama prechki na nivo design, koito da go pravjat nevuzmozhno > > On Wednesday 05 October 2005 09:07, Ivan Peikov wrote: > > razbira se. bil sum pijan kato sum go pisal :) shte go slozha da > > destroyva predi da go initializira. > > > > po problema sus transition explosion-a. > > vchera mislih za edno reshenie, koeto ni spasjava zadnicite, obache pravi > > avtomatite malko po-trudni za programirane. > > > > reshenieto e prosto - vmesto simvol v transition-ite shte pazim range ot > > simvoli. tova shte iziskva tochno oshte edin byte. izobshto ako simvol ot > > azbukata se predstavja s X byte-a ni trjabvat tochno oshte X byte-a za da > > predstavim range ot simvoli (v purvite X pishem base simvola na range-a, > > vuv vtorite po-izbor - ili kraja na range-a ili broja simvoli vlizashti v > > range-a sled base) > > > > s kakvo tova prestavjane e po-dobro. ami dokazuemo e che transitionite > > polucheni pri konstruiraneto na NFA (po tazi shema) sa <= broja na > > simvolite v regular expression-a ako njama +/*. +/* operaciite vdigat na > > kvadrat broja na transitionite v avtomata kum kojto sa prilozheni (ako > > trjabva da sum tochen, uvelichava s broja na krajnite po broja na > > nachalnite, koeto e po-malko, no pak e ot porjaduka na O(n^2)). > > specialno v sluchaja za kojto spomenah v prednija mail ".*", "." generira > > 2 sustojanija i slaga transition s range 0-255 na nego. sled tova > > zvezdata dobavja edno krajno sustojanie ot koeto s nishto ne se izliza > > (za da machne epsilon) i dobavja cikul ot nachalnoto kum nego si pak s > > label 0-255. toest transitionite stavat 2, a sustojanijata 3 - dosta > > po-dobre ot 9120 kum 191.. tova predstavjane sushto taka e dobro za > > predstavjane na clasovete ot simvoli - neshta kato [[:alpha:]0-9], se > > predstavjat super trivialno s ranges. > > > > loshoto na range-ovete e che pravjat implementacijata malko po-trudna. > > predstavjaneto veche ne tochno tova koeto chovek ima predvid kato si > > misli za graph/avtomat. predpolagam obache che shte trjabva da zhiveem s > > tova, zashtoto v purvonachalnija variant avtomatite sa absolutno > > neizpolzvaemi.. > > > > mnenija, drugi idei? > > > > On Tuesday 04 October 2005 18:14, Georgi Mungov wrote: > > > rex_parser_parse ne e li dobre da si destroy-va scanner-a nakraia ili > > > da destroy-va staria v nachaloto. sega ako napravish 2 puti > > > > > > rex_parser_parse() > > > rex_parser_parse() > > > > > > shte leak-nesh > > > > > > On Tuesday 04 October 2005 09:58, Ivan Peikov wrote: > > > > Pichove eto posledna versija. > > > > Dnes po njakoe vreme shte ja slozha v CVS-a na SF.net za da mozhe da > > > > se raboti po-normalno. > > > > > > > > Kakvo imame tuk: > > > > - optimizacii po regex parse durvoto > > > > - implementacija GFA/NFA > > > > > > > > generirame NFA ot reguljaren izraz. vse oshte ne se poddurzhat > > > > klasove ([...]), pochti sum gi napravil, ama ne suvsem i za tova ne > > > > sum gi vkljuchil tuk. mozhete da izprobvate tools/nfa [REGEX] - dumpi > > > > v dot format generiranija avtomat. demek neshto ot sorta bi trjabvalo > > > > da go vizualizira: > > > > > > > > nfa "a+(ab|bc)c+" | dot -Tps -oshit.ps ; kghostview shit.ps > > > > > > > > nachi kato cjalo vsichko e pushka, imam obache slednite neprijatni > > > > nabljudenija. reguljarnijat izraz ".*" napravo razcepva mraka. dori > > > > kogato interpretiram '.' kato "vsichki printable characters" > > > > generiranijat avtomat ima 191 sustojanija i 9120 transition-a!! koeto > > > > na 32-bit os oznachava grubo 86K za nedeterminiranija avtomat. ne > > > > znam kolko mozhe da narastne tova pri determiniranija (principno > > > > mozhe da narastne exponencialno...) > > > > > > > > tova naj-malkoto objasnjava zashto nikoj ne implementira reguljarni > > > > izrazi s krajni avtomati :) > > > > > > > > ne sum mnogo siguren che trjabva da reshavame tozi problem (verojatno > > > > trjabva ako iskame njakoj da izpolzva izobshto libfsd za drugo osven > > > > za demonstracionni celi), no njakakvo reshenie e da pishem v > > > > transition-a tablica ot simvoli, a ne samo edin simvol. > > > > > > > > shte vidja kakvo kazvat Aho i kolegija v knigata Compilers. maj > > > > imashe neshto za predstavjane na krajni avtomati (algorithm na Tarjan > > > > ili neshto takova). > > > > > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > > > annotations on 10/04/05 08:23:49 > > > ----------------------------------------------------------------------- > > >-- -- --- This e-mail, including attachments, may include confidential > > > and/or proprietary information, and may be used only by the person or > > > entity to which it is addressed. If the reader of this e-mail is not > > > the intended recipient or his or her authorized agent, the reader is > > > hereby notified that any dissemination, distribution or copying of this > > > e-mail is prohibited. If you have received this e-mail in error, please > > > notify the sender by replying to this message and delete this e-mail > > > immediately. > > > ======================================================================= > > >== == === > > > > > > > > > > > > ------------------------------------------------------- > > > This SF.Net email is sponsored by: > > > Power Architecture Resource Center: Free content, downloads, > > > discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl > > > _______________________________________________ > > > Libfsd-devel mailing list > > > Lib...@li... > > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: > > Power Architecture Resource Center: Free content, downloads, discussions, > > and more. http://solutions.newsforge.com/ibmarch.tmpl > > _______________________________________________ > > Libfsd-devel mailing list > > Lib...@li... > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > annotations on 10/05/05 00:42:16 > --------------------------------------------------------------------------- >--- This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified > that any dissemination, distribution or copying of this e-mail is > prohibited. If you have received this e-mail in error, please notify the > sender by replying to this message and delete this e-mail immediately. > =========================================================================== >=== > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel |
|
From: Georgi M. <geo...@tu...> - 2005-10-05 07:33:54
|
Moje bi po skoro e list ot ranges zashtoto [0-9A-Za-z] sa realno 3 range-a No togava mai shte stane mnogo tromavo. Vidiah che si fixiral azbukata na ASCII. Mojem li da predstavim togava range-a (ili ranges) sus bitset On Wednesday 05 October 2005 09:07, Ivan Peikov wrote: > razbira se. bil sum pijan kato sum go pisal :) shte go slozha da destroyva > predi da go initializira. > > po problema sus transition explosion-a. > vchera mislih za edno reshenie, koeto ni spasjava zadnicite, obache pravi > avtomatite malko po-trudni za programirane. > > reshenieto e prosto - vmesto simvol v transition-ite shte pazim range ot > simvoli. tova shte iziskva tochno oshte edin byte. izobshto ako simvol ot > azbukata se predstavja s X byte-a ni trjabvat tochno oshte X byte-a za da > predstavim range ot simvoli (v purvite X pishem base simvola na range-a, > vuv vtorite po-izbor - ili kraja na range-a ili broja simvoli vlizashti v > range-a sled base) > > s kakvo tova prestavjane e po-dobro. ami dokazuemo e che transitionite > polucheni pri konstruiraneto na NFA (po tazi shema) sa <= broja na > simvolite v regular expression-a ako njama +/*. +/* operaciite vdigat na > kvadrat broja na transitionite v avtomata kum kojto sa prilozheni (ako > trjabva da sum tochen, uvelichava s broja na krajnite po broja na > nachalnite, koeto e po-malko, no pak e ot porjaduka na O(n^2)). > specialno v sluchaja za kojto spomenah v prednija mail ".*", "." generira 2 > sustojanija i slaga transition s range 0-255 na nego. sled tova zvezdata > dobavja edno krajno sustojanie ot koeto s nishto ne se izliza (za da machne > epsilon) i dobavja cikul ot nachalnoto kum nego si pak s label 0-255. toest > transitionite stavat 2, a sustojanijata 3 - dosta po-dobre ot 9120 kum > 191.. tova predstavjane sushto taka e dobro za predstavjane na clasovete ot > simvoli - neshta kato [[:alpha:]0-9], se predstavjat super trivialno s > ranges. > > loshoto na range-ovete e che pravjat implementacijata malko po-trudna. > predstavjaneto veche ne tochno tova koeto chovek ima predvid kato si misli > za graph/avtomat. predpolagam obache che shte trjabva da zhiveem s tova, > zashtoto v purvonachalnija variant avtomatite sa absolutno neizpolzvaemi.. > > mnenija, drugi idei? > > On Tuesday 04 October 2005 18:14, Georgi Mungov wrote: > > rex_parser_parse ne e li dobre da si destroy-va scanner-a nakraia ili da > > destroy-va staria v nachaloto. sega ako napravish 2 puti > > > > rex_parser_parse() > > rex_parser_parse() > > > > shte leak-nesh > > > > On Tuesday 04 October 2005 09:58, Ivan Peikov wrote: > > > Pichove eto posledna versija. > > > Dnes po njakoe vreme shte ja slozha v CVS-a na SF.net za da mozhe da se > > > raboti po-normalno. > > > > > > Kakvo imame tuk: > > > - optimizacii po regex parse durvoto > > > - implementacija GFA/NFA > > > > > > generirame NFA ot reguljaren izraz. vse oshte ne se poddurzhat klasove > > > ([...]), pochti sum gi napravil, ama ne suvsem i za tova ne sum gi > > > vkljuchil tuk. mozhete da izprobvate tools/nfa [REGEX] - dumpi v dot > > > format generiranija avtomat. demek neshto ot sorta bi trjabvalo da go > > > vizualizira: > > > > > > nfa "a+(ab|bc)c+" | dot -Tps -oshit.ps ; kghostview shit.ps > > > > > > nachi kato cjalo vsichko e pushka, imam obache slednite neprijatni > > > nabljudenija. reguljarnijat izraz ".*" napravo razcepva mraka. dori > > > kogato interpretiram '.' kato "vsichki printable characters" > > > generiranijat avtomat ima 191 sustojanija i 9120 transition-a!! koeto > > > na 32-bit os oznachava grubo 86K za nedeterminiranija avtomat. ne znam > > > kolko mozhe da narastne tova pri determiniranija (principno mozhe da > > > narastne exponencialno...) > > > > > > tova naj-malkoto objasnjava zashto nikoj ne implementira reguljarni > > > izrazi s krajni avtomati :) > > > > > > ne sum mnogo siguren che trjabva da reshavame tozi problem (verojatno > > > trjabva ako iskame njakoj da izpolzva izobshto libfsd za drugo osven za > > > demonstracionni celi), no njakakvo reshenie e da pishem v transition-a > > > tablica ot simvoli, a ne samo edin simvol. > > > > > > shte vidja kakvo kazvat Aho i kolegija v knigata Compilers. maj imashe > > > neshto za predstavjane na krajni avtomati (algorithm na Tarjan ili > > > neshto takova). > > > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > > annotations on 10/04/05 08:23:49 > > ------------------------------------------------------------------------- > >-- --- This e-mail, including attachments, may include confidential and/or > > proprietary information, and may be used only by the person or entity to > > which it is addressed. If the reader of this e-mail is not the intended > > recipient or his or her authorized agent, the reader is hereby notified > > that any dissemination, distribution or copying of this e-mail is > > prohibited. If you have received this e-mail in error, please notify the > > sender by replying to this message and delete this e-mail immediately. > > ========================================================================= > >== === > > > > > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: > > Power Architecture Resource Center: Free content, downloads, discussions, > > and more. http://solutions.newsforge.com/ibmarch.tmpl > > _______________________________________________ > > Libfsd-devel mailing list > > Lib...@li... > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 10/05/05 00:42:16 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-10-05 06:08:05
|
razbira se. bil sum pijan kato sum go pisal :) shte go slozha da destroyva predi da go initializira. po problema sus transition explosion-a. vchera mislih za edno reshenie, koeto ni spasjava zadnicite, obache pravi avtomatite malko po-trudni za programirane. reshenieto e prosto - vmesto simvol v transition-ite shte pazim range ot simvoli. tova shte iziskva tochno oshte edin byte. izobshto ako simvol ot azbukata se predstavja s X byte-a ni trjabvat tochno oshte X byte-a za da predstavim range ot simvoli (v purvite X pishem base simvola na range-a, vuv vtorite po-izbor - ili kraja na range-a ili broja simvoli vlizashti v range-a sled base) s kakvo tova prestavjane e po-dobro. ami dokazuemo e che transitionite polucheni pri konstruiraneto na NFA (po tazi shema) sa <= broja na simvolite v regular expression-a ako njama +/*. +/* operaciite vdigat na kvadrat broja na transitionite v avtomata kum kojto sa prilozheni (ako trjabva da sum tochen, uvelichava s broja na krajnite po broja na nachalnite, koeto e po-malko, no pak e ot porjaduka na O(n^2)). specialno v sluchaja za kojto spomenah v prednija mail ".*", "." generira 2 sustojanija i slaga transition s range 0-255 na nego. sled tova zvezdata dobavja edno krajno sustojanie ot koeto s nishto ne se izliza (za da machne epsilon) i dobavja cikul ot nachalnoto kum nego si pak s label 0-255. toest transitionite stavat 2, a sustojanijata 3 - dosta po-dobre ot 9120 kum 191.. tova predstavjane sushto taka e dobro za predstavjane na clasovete ot simvoli - neshta kato [[:alpha:]0-9], se predstavjat super trivialno s ranges. loshoto na range-ovete e che pravjat implementacijata malko po-trudna. predstavjaneto veche ne tochno tova koeto chovek ima predvid kato si misli za graph/avtomat. predpolagam obache che shte trjabva da zhiveem s tova, zashtoto v purvonachalnija variant avtomatite sa absolutno neizpolzvaemi.. mnenija, drugi idei? On Tuesday 04 October 2005 18:14, Georgi Mungov wrote: > rex_parser_parse ne e li dobre da si destroy-va scanner-a nakraia ili da > destroy-va staria v nachaloto. sega ako napravish 2 puti > > rex_parser_parse() > rex_parser_parse() > > shte leak-nesh > > On Tuesday 04 October 2005 09:58, Ivan Peikov wrote: > > Pichove eto posledna versija. > > Dnes po njakoe vreme shte ja slozha v CVS-a na SF.net za da mozhe da se > > raboti po-normalno. > > > > Kakvo imame tuk: > > - optimizacii po regex parse durvoto > > - implementacija GFA/NFA > > > > generirame NFA ot reguljaren izraz. vse oshte ne se poddurzhat klasove > > ([...]), pochti sum gi napravil, ama ne suvsem i za tova ne sum gi > > vkljuchil tuk. mozhete da izprobvate tools/nfa [REGEX] - dumpi v dot > > format generiranija avtomat. demek neshto ot sorta bi trjabvalo da go > > vizualizira: > > > > nfa "a+(ab|bc)c+" | dot -Tps -oshit.ps ; kghostview shit.ps > > > > nachi kato cjalo vsichko e pushka, imam obache slednite neprijatni > > nabljudenija. reguljarnijat izraz ".*" napravo razcepva mraka. dori > > kogato interpretiram '.' kato "vsichki printable characters" > > generiranijat avtomat ima 191 sustojanija i 9120 transition-a!! koeto na > > 32-bit os oznachava grubo 86K za nedeterminiranija avtomat. ne znam kolko > > mozhe da narastne tova pri determiniranija (principno mozhe da narastne > > exponencialno...) > > > > tova naj-malkoto objasnjava zashto nikoj ne implementira reguljarni > > izrazi s krajni avtomati :) > > > > ne sum mnogo siguren che trjabva da reshavame tozi problem (verojatno > > trjabva ako iskame njakoj da izpolzva izobshto libfsd za drugo osven za > > demonstracionni celi), no njakakvo reshenie e da pishem v transition-a > > tablica ot simvoli, a ne samo edin simvol. > > > > shte vidja kakvo kazvat Aho i kolegija v knigata Compilers. maj imashe > > neshto za predstavjane na krajni avtomati (algorithm na Tarjan ili neshto > > takova). > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > annotations on 10/04/05 08:23:49 > --------------------------------------------------------------------------- >--- This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified > that any dissemination, distribution or copying of this e-mail is > prohibited. If you have received this e-mail in error, please notify the > sender by replying to this message and delete this e-mail immediately. > =========================================================================== >=== > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel |
|
From: Georgi M. <geo...@tu...> - 2005-10-04 15:15:40
|
rex_parser_parse ne e li dobre da si destroy-va scanner-a nakraia ili da destroy-va staria v nachaloto. sega ako napravish 2 puti rex_parser_parse() rex_parser_parse() shte leak-nesh On Tuesday 04 October 2005 09:58, Ivan Peikov wrote: > Pichove eto posledna versija. > Dnes po njakoe vreme shte ja slozha v CVS-a na SF.net za da mozhe da se > raboti po-normalno. > > Kakvo imame tuk: > - optimizacii po regex parse durvoto > - implementacija GFA/NFA > > generirame NFA ot reguljaren izraz. vse oshte ne se poddurzhat klasove > ([...]), pochti sum gi napravil, ama ne suvsem i za tova ne sum gi > vkljuchil tuk. mozhete da izprobvate tools/nfa [REGEX] - dumpi v dot format > generiranija avtomat. demek neshto ot sorta bi trjabvalo da go vizualizira: > > nfa "a+(ab|bc)c+" | dot -Tps -oshit.ps ; kghostview shit.ps > > nachi kato cjalo vsichko e pushka, imam obache slednite neprijatni > nabljudenija. reguljarnijat izraz ".*" napravo razcepva mraka. dori kogato > interpretiram '.' kato "vsichki printable characters" generiranijat avtomat > ima 191 sustojanija i 9120 transition-a!! koeto na 32-bit os oznachava > grubo 86K za nedeterminiranija avtomat. ne znam kolko mozhe da narastne > tova pri determiniranija (principno mozhe da narastne exponencialno...) > > tova naj-malkoto objasnjava zashto nikoj ne implementira reguljarni izrazi > s krajni avtomati :) > > ne sum mnogo siguren che trjabva da reshavame tozi problem (verojatno > trjabva ako iskame njakoj da izpolzva izobshto libfsd za drugo osven za > demonstracionni celi), no njakakvo reshenie e da pishem v transition-a > tablica ot simvoli, a ne samo edin simvol. > > shte vidja kakvo kazvat Aho i kolegija v knigata Compilers. maj imashe > neshto za predstavjane na krajni avtomati (algorithm na Tarjan ili neshto > takova). "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 10/04/05 08:23:49 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-10-04 06:58:35
|
Pichove eto posledna versija. Dnes po njakoe vreme shte ja slozha v CVS-a na SF.net za da mozhe da se raboti po-normalno. Kakvo imame tuk: - optimizacii po regex parse durvoto - implementacija GFA/NFA generirame NFA ot reguljaren izraz. vse oshte ne se poddurzhat klasove ([...]), pochti sum gi napravil, ama ne suvsem i za tova ne sum gi vkljuchil tuk. mozhete da izprobvate tools/nfa [REGEX] - dumpi v dot format generiranija avtomat. demek neshto ot sorta bi trjabvalo da go vizualizira: nfa "a+(ab|bc)c+" | dot -Tps -oshit.ps ; kghostview shit.ps nachi kato cjalo vsichko e pushka, imam obache slednite neprijatni nabljudenija. reguljarnijat izraz ".*" napravo razcepva mraka. dori kogato interpretiram '.' kato "vsichki printable characters" generiranijat avtomat ima 191 sustojanija i 9120 transition-a!! koeto na 32-bit os oznachava grubo 86K za nedeterminiranija avtomat. ne znam kolko mozhe da narastne tova pri determiniranija (principno mozhe da narastne exponencialno...) tova naj-malkoto objasnjava zashto nikoj ne implementira reguljarni izrazi s krajni avtomati :) ne sum mnogo siguren che trjabva da reshavame tozi problem (verojatno trjabva ako iskame njakoj da izpolzva izobshto libfsd za drugo osven za demonstracionni celi), no njakakvo reshenie e da pishem v transition-a tablica ot simvoli, a ne samo edin simvol. shte vidja kakvo kazvat Aho i kolegija v knigata Compilers. maj imashe neshto za predstavjane na krajni avtomati (algorithm na Tarjan ili neshto takova). |
|
From: Ivan P. <va...@ne...> - 2005-10-03 11:39:35
|
On Monday 03 October 2005 14:03, Georgi Mungov wrote: > niama li da e dobre da davame na fsd_pool-a niakuv allocator. niakoi moje mislih za tova i reshih che mnogo malko verojatno e njakoj da poiska da si alokira avtomat v shared memory :) principno si prav. naj-veche bi bilo gotino da imame za autotesta nachin da dadem debug allocator, kojto da testva, a za lib-a da se podava NULL i da si polzva heap allocatora. taka za razlika ot NC-libs njakoj mozhe da autotestva s extended memory checks dazhe da ne e kompiliral s -DDEBUG > da iska s shared memory da go polzva. i ima li smisal ot zalloc - calloc > sushto ti nulira pametta > tova koeto moga da napravja e #define-na ZALLOC(x) da e calloc(1, x), no kato cjalo iskah neshto kato malloc, ama da mi nulira, a ne mi se pishe tazi edinica navsjakude > probvah da compile-na na 64bit > > ============== > rex_parser.c: In function `rex_node_dump_r': > rex_parser.c:559: warning: int format, different type arg (arg 6) > rex_parser.c:559: warning: int format, different type arg (arg 7) > ================= > > ================= > Making all in tools > mkdir .libs > gcc -Wall -Werror -g2 -DDEBUG -g -O2 -o rex rex.o ../src/.libs/libfsd.al > gfa.c: In function `main': > gfa.c:14: warning: cast to pointer from integer of different size > gfa.c:19: warning: cast from pointer to integer of different size > gfa.c:20: warning: cast from pointer to integer of different size > gfa.c:30: warning: cast from pointer to integer of different size > ================= > > void* i size_t po princip sa long a ne int > ok, tuj kato njamam 64-bit linux vkushti, naznachavame te otgovornik po 64-bitovija build :) > On Monday 03 October 2005 11:46, Ivan Peikov wrote: > > Wow!! Kak bjaha NightWish? Kosta ne hodi li (vidjah go v subota vecher na > > partito na Dara puk nishto ne spomena)? > > > > Inache az utre shte pratja versija s mnogo promeni po neja (optimized > > regular expressions + GFA + NFA) > > > > A NC-Base AT mozhe naj-malko da pitame Lenkov - ne mislja che shte se > > zapune. Toj i bez drugo shte gi puska (L)GPL tija libs po njakoe vreme > > > > Utre kato vi pratja poslednoto distro mislja da go commitna v CVS-a ns > > SF.net zashtoto veche stava trudno da pravim neshto paralelno. Taka che > > ako imate neshto zabelezhki po strukturata kazvajte sega, che posle ako > > iskame da rename-vame/iztrivame neshto shte trjabva da iskame ot SF > > support-a. > > > > Attachvam snimka na purvija NFA generiran ot libfsd (istoricheski > > dokument) ot reguljarnija izraz "(ab|bc)xx" :) > > > > On Monday 03 October 2005 10:25, Georgi Mungov wrote: > > > Dovechera shte go vidia i az vsichkoto che petuk i subota biah v > > > bukuresht na koncert na nightwish > > > > > > Za AT ot NC-Base -- toi svobodno dostupen li e za vsichki? > > > > > > On Friday 30 September 2005 13:03, Ivan Peikov wrote: > > > > Za autotest framework-a, misleh si che ako go dobavim v distroto shte > > > > go slozhim otdelno v libfsd/autotest foldera i njama nito da se > > > > instalira nito da se nabiva na ochi. ot druga strana bi ni spestilo > > > > razni neshta (memory leak test naprimer kakto i formatting-a na > > > > rezultatite. Demek sus sigurnost shte si spestim njakakva rabota. > > > > Mezhdu drugoto, sega useshtam che s tezi pool-ove testvaneto za > > > > leakove njama da e tolkova prosto - kato cjalo leakove njama da ima, > > > > no mozhem da leakvame na nivo pool, koeto pak e zle. > > > > > > > > Za collating sequences i equivallence classes (osobeno za purvoto), > > > > baja se muchih i maj ne razbrah napulno kakva e idejata. Vse pak > > > > attachvam edin html kojto gi specificira tija fenomeni. Sushto taka > > > > perllocale e dosta polezen - tam sa objasneni s po-choveshki dumi. > > > > Drugoto e re.html, koeto e v libfsd/doc -- tam e gramatikata na > > > > extended reguljaren izraz. > > > > > > > > On Friday 30 September 2005 12:51, Petar Petrov wrote: > > > > > Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. > > > > > Gledam, che si imame i lexer (rex_scanner.c). > > > > > > > > > > Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C > > > > > code. Ako vzemem code-a ot at-framework-a shte trqbva da go dobavim > > > > > v libfsd distroto, a tova shte e ~40% ot celiq code. Koeto e malko > > > > > typo. > > > > > > > > > > Shte vidq dali shte moga da dobavq collating sequences i > > > > > equivalence classes v reg ex. parser-a. > > > > > > > > > > On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > > > > > Prashtam distroto dokudeto sum go dobutal. > > > > > > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti > > > > > > i sushtevremenno lesno da se programirat i nakraja napisah edin > > > > > > super generic > > > > > > ciklichen single-linked list, podoben na NC-Base list-a, obache > > > > > > jade po-malko > > > > > > pamet i polzva pool za da si allocira node-ovete. Pool-a se > > > > > > podava i globalen > > > > > > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > > > > > > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se > > > > > > alocira pamet za > > > > > > nego. Tova e za da se embedva v drugi strukturi, za koito ima > > > > > > pool (state_t i > > > > > > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s > > > > > > malloc e samijat avtomat. > > > > > > Pesho - rex parser-a - nadjavam se da se izkefish :) > > > > > > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato > > > > > > kovem osnovite > > > > > > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto > > > > > > da e dostatuchno generic taka che da ne zavisim ot neshto, koeto > > > > > > trjabva da se instalira (neshto koeto go njama vuv vsjaka moderna > > > > > > linux distribucija). Variant e na perl test frameworka primerno. > > > > > > No trjabva neshto da se izmisli > > > > > > za da si pishem testovete zaedno s koda. Inache shte e mnogo > > > > > > trudno posle da > > > > > > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama > > > > > > da go napravim izobshto) a mi se iska vse pak da imame njakakuv > > > > > > nachin da si podsigurim kachestvoto na lib-a inache ne si struva > > > > > > zagubenoto vreme. Sushto variant e da si vzemem ot NC-Base > > > > > > at_framework-a i da go nagodim (i bez > > > > > > drugo nie sme si go mislili i sme sviknali da si go pozvame) > > > > > > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto > > > > > > vidite kato > > > > > > problem -- kazvajte. > > > > > > > > > > > > Ajde leka i pishete. > > > > > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > > > annotations on 10/03/05 00:34:11 > > > ----------------------------------------------------------------------- > > >-- -- --- This e-mail, including attachments, may include confidential > > > and/or proprietary information, and may be used only by the person or > > > entity to which it is addressed. If the reader of this e-mail is not > > > the intended recipient or his or her authorized agent, the reader is > > > hereby notified that any dissemination, distribution or copying of this > > > e-mail is prohibited. If you have received this e-mail in error, please > > > notify the sender by replying to this message and delete this e-mail > > > immediately. > > > ======================================================================= > > >== == === > > > > > > > > > > > > ------------------------------------------------------- > > > This SF.Net email is sponsored by: > > > Power Architecture Resource Center: Free content, downloads, > > > discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl > > > _______________________________________________ > > > Libfsd-devel mailing list > > > Lib...@li... > > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > annotations on 10/03/05 04:12:44 > --------------------------------------------------------------------------- >--- This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified > that any dissemination, distribution or copying of this e-mail is > prohibited. If you have received this e-mail in error, please notify the > sender by replying to this message and delete this e-mail immediately. > =========================================================================== >=== > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel |
|
From: Georgi M. <geo...@tu...> - 2005-10-03 11:04:25
|
niama li da e dobre da davame na fsd_pool-a niakuv allocator. niakoi moje da iska s shared memory da go polzva. i ima li smisal ot zalloc - calloc sushto ti nulira pametta probvah da compile-na na 64bit ============== rex_parser.c: In function `rex_node_dump_r': rex_parser.c:559: warning: int format, different type arg (arg 6) rex_parser.c:559: warning: int format, different type arg (arg 7) ================= ================= Making all in tools mkdir .libs gcc -Wall -Werror -g2 -DDEBUG -g -O2 -o rex rex.o ../src/.libs/libfsd.al gfa.c: In function `main': gfa.c:14: warning: cast to pointer from integer of different size gfa.c:19: warning: cast from pointer to integer of different size gfa.c:20: warning: cast from pointer to integer of different size gfa.c:30: warning: cast from pointer to integer of different size ================= void* i size_t po princip sa long a ne int On Monday 03 October 2005 11:46, Ivan Peikov wrote: > Wow!! Kak bjaha NightWish? Kosta ne hodi li (vidjah go v subota vecher na > partito na Dara puk nishto ne spomena)? > > Inache az utre shte pratja versija s mnogo promeni po neja (optimized > regular expressions + GFA + NFA) > > A NC-Base AT mozhe naj-malko da pitame Lenkov - ne mislja che shte se > zapune. Toj i bez drugo shte gi puska (L)GPL tija libs po njakoe vreme > > Utre kato vi pratja poslednoto distro mislja da go commitna v CVS-a ns > SF.net zashtoto veche stava trudno da pravim neshto paralelno. Taka che ako > imate neshto zabelezhki po strukturata kazvajte sega, che posle ako iskame > da rename-vame/iztrivame neshto shte trjabva da iskame ot SF support-a. > > Attachvam snimka na purvija NFA generiran ot libfsd (istoricheski dokument) > ot reguljarnija izraz "(ab|bc)xx" :) > > On Monday 03 October 2005 10:25, Georgi Mungov wrote: > > Dovechera shte go vidia i az vsichkoto che petuk i subota biah v > > bukuresht na koncert na nightwish > > > > Za AT ot NC-Base -- toi svobodno dostupen li e za vsichki? > > > > On Friday 30 September 2005 13:03, Ivan Peikov wrote: > > > Za autotest framework-a, misleh si che ako go dobavim v distroto shte > > > go slozhim otdelno v libfsd/autotest foldera i njama nito da se > > > instalira nito da se nabiva na ochi. ot druga strana bi ni spestilo > > > razni neshta (memory leak test naprimer kakto i formatting-a na > > > rezultatite. Demek sus sigurnost shte si spestim njakakva rabota. > > > Mezhdu drugoto, sega useshtam che s tezi pool-ove testvaneto za leakove > > > njama da e tolkova prosto - kato cjalo leakove njama da ima, no mozhem > > > da leakvame na nivo pool, koeto pak e zle. > > > > > > Za collating sequences i equivallence classes (osobeno za purvoto), > > > baja se muchih i maj ne razbrah napulno kakva e idejata. Vse pak > > > attachvam edin html kojto gi specificira tija fenomeni. Sushto taka > > > perllocale e dosta polezen - tam sa objasneni s po-choveshki dumi. > > > Drugoto e re.html, koeto e v libfsd/doc -- tam e gramatikata na > > > extended reguljaren izraz. > > > > > > On Friday 30 September 2005 12:51, Petar Petrov wrote: > > > > Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. > > > > Gledam, che si imame i lexer (rex_scanner.c). > > > > > > > > Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C > > > > code. Ako vzemem code-a ot at-framework-a shte trqbva da go dobavim v > > > > libfsd distroto, a tova shte e ~40% ot celiq code. Koeto e malko > > > > typo. > > > > > > > > Shte vidq dali shte moga da dobavq collating sequences i equivalence > > > > classes v reg ex. parser-a. > > > > > > > > On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > > > > Prashtam distroto dokudeto sum go dobutal. > > > > > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i > > > > > sushtevremenno lesno da se programirat i nakraja napisah edin super > > > > > generic > > > > > ciklichen single-linked list, podoben na NC-Base list-a, obache > > > > > jade po-malko > > > > > pamet i polzva pool za da si allocira node-ovete. Pool-a se podava > > > > > i globalen > > > > > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > > > > > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se > > > > > alocira pamet za > > > > > nego. Tova e za da se embedva v drugi strukturi, za koito ima pool > > > > > (state_t i > > > > > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s > > > > > malloc e samijat avtomat. > > > > > Pesho - rex parser-a - nadjavam se da se izkefish :) > > > > > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem > > > > > osnovite > > > > > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto > > > > > da e dostatuchno generic taka che da ne zavisim ot neshto, koeto > > > > > trjabva da se instalira (neshto koeto go njama vuv vsjaka moderna > > > > > linux distribucija). Variant e na perl test frameworka primerno. No > > > > > trjabva neshto da se izmisli > > > > > za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno > > > > > posle da > > > > > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama > > > > > da go napravim izobshto) a mi se iska vse pak da imame njakakuv > > > > > nachin da si podsigurim kachestvoto na lib-a inache ne si struva > > > > > zagubenoto vreme. Sushto variant e da si vzemem ot NC-Base > > > > > at_framework-a i da go nagodim (i bez > > > > > drugo nie sme si go mislili i sme sviknali da si go pozvame) > > > > > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto > > > > > vidite kato > > > > > problem -- kazvajte. > > > > > > > > > > Ajde leka i pishete. > > > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > > annotations on 10/03/05 00:34:11 > > ------------------------------------------------------------------------- > >-- --- This e-mail, including attachments, may include confidential and/or > > proprietary information, and may be used only by the person or entity to > > which it is addressed. If the reader of this e-mail is not the intended > > recipient or his or her authorized agent, the reader is hereby notified > > that any dissemination, distribution or copying of this e-mail is > > prohibited. If you have received this e-mail in error, please notify the > > sender by replying to this message and delete this e-mail immediately. > > ========================================================================= > >== === > > > > > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: > > Power Architecture Resource Center: Free content, downloads, discussions, > > and more. http://solutions.newsforge.com/ibmarch.tmpl > > _______________________________________________ > > Libfsd-devel mailing list > > Lib...@li... > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 10/03/05 04:12:44 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-10-03 08:47:54
|
Wow!! Kak bjaha NightWish? Kosta ne hodi li (vidjah go v subota vecher na partito na Dara puk nishto ne spomena)? Inache az utre shte pratja versija s mnogo promeni po neja (optimized regular expressions + GFA + NFA) A NC-Base AT mozhe naj-malko da pitame Lenkov - ne mislja che shte se zapune. Toj i bez drugo shte gi puska (L)GPL tija libs po njakoe vreme Utre kato vi pratja poslednoto distro mislja da go commitna v CVS-a ns SF.net zashtoto veche stava trudno da pravim neshto paralelno. Taka che ako imate neshto zabelezhki po strukturata kazvajte sega, che posle ako iskame da rename-vame/iztrivame neshto shte trjabva da iskame ot SF support-a. Attachvam snimka na purvija NFA generiran ot libfsd (istoricheski dokument) ot reguljarnija izraz "(ab|bc)xx" :) On Monday 03 October 2005 10:25, Georgi Mungov wrote: > Dovechera shte go vidia i az vsichkoto che petuk i subota biah v bukuresht > na koncert na nightwish > > Za AT ot NC-Base -- toi svobodno dostupen li e za vsichki? > > On Friday 30 September 2005 13:03, Ivan Peikov wrote: > > Za autotest framework-a, misleh si che ako go dobavim v distroto shte go > > slozhim otdelno v libfsd/autotest foldera i njama nito da se instalira > > nito da se nabiva na ochi. ot druga strana bi ni spestilo razni neshta > > (memory leak test naprimer kakto i formatting-a na rezultatite. Demek sus > > sigurnost shte si spestim njakakva rabota. > > Mezhdu drugoto, sega useshtam che s tezi pool-ove testvaneto za leakove > > njama da e tolkova prosto - kato cjalo leakove njama da ima, no mozhem da > > leakvame na nivo pool, koeto pak e zle. > > > > Za collating sequences i equivallence classes (osobeno za purvoto), baja > > se muchih i maj ne razbrah napulno kakva e idejata. Vse pak attachvam > > edin html kojto gi specificira tija fenomeni. Sushto taka perllocale e > > dosta polezen - tam sa objasneni s po-choveshki dumi. Drugoto e re.html, > > koeto e v libfsd/doc -- tam e gramatikata na extended reguljaren izraz. > > > > On Friday 30 September 2005 12:51, Petar Petrov wrote: > > > Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. Gledam, > > > che si imame i lexer (rex_scanner.c). > > > > > > Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C code. > > > Ako vzemem code-a ot at-framework-a shte trqbva da go dobavim v libfsd > > > distroto, a tova shte e ~40% ot celiq code. Koeto e malko typo. > > > > > > Shte vidq dali shte moga da dobavq collating sequences i equivalence > > > classes v reg ex. parser-a. > > > > > > On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > > > Prashtam distroto dokudeto sum go dobutal. > > > > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i > > > > sushtevremenno lesno da se programirat i nakraja napisah edin super > > > > generic > > > > ciklichen single-linked list, podoben na NC-Base list-a, obache jade > > > > po-malko > > > > pamet i polzva pool za da si allocira node-ovete. Pool-a se podava i > > > > globalen > > > > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > > > > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se alocira > > > > pamet za > > > > nego. Tova e za da se embedva v drugi strukturi, za koito ima pool > > > > (state_t i > > > > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s > > > > malloc e samijat avtomat. > > > > Pesho - rex parser-a - nadjavam se da se izkefish :) > > > > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem > > > > osnovite > > > > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto da > > > > e dostatuchno generic taka che da ne zavisim ot neshto, koeto trjabva > > > > da se instalira (neshto koeto go njama vuv vsjaka moderna linux > > > > distribucija). Variant e na perl test frameworka primerno. No trjabva > > > > neshto da se izmisli > > > > za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno > > > > posle da > > > > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama da > > > > go napravim izobshto) a mi se iska vse pak da imame njakakuv nachin > > > > da si podsigurim kachestvoto na lib-a inache ne si struva zagubenoto > > > > vreme. Sushto variant e da si vzemem ot NC-Base at_framework-a i da > > > > go nagodim (i bez > > > > drugo nie sme si go mislili i sme sviknali da si go pozvame) > > > > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto > > > > vidite kato > > > > problem -- kazvajte. > > > > > > > > Ajde leka i pishete. > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following > annotations on 10/03/05 00:34:11 > --------------------------------------------------------------------------- >--- This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified > that any dissemination, distribution or copying of this e-mail is > prohibited. If you have received this e-mail in error, please notify the > sender by replying to this message and delete this e-mail immediately. > =========================================================================== >=== > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: > Power Architecture Resource Center: Free content, downloads, discussions, > and more. http://solutions.newsforge.com/ibmarch.tmpl > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel |
|
From: Georgi M. <geo...@tu...> - 2005-10-03 07:25:56
|
Dovechera shte go vidia i az vsichkoto che petuk i subota biah v bukuresht na koncert na nightwish Za AT ot NC-Base -- toi svobodno dostupen li e za vsichki? On Friday 30 September 2005 13:03, Ivan Peikov wrote: > Za autotest framework-a, misleh si che ako go dobavim v distroto shte go > slozhim otdelno v libfsd/autotest foldera i njama nito da se instalira nito > da se nabiva na ochi. ot druga strana bi ni spestilo razni neshta (memory > leak test naprimer kakto i formatting-a na rezultatite. Demek sus sigurnost > shte si spestim njakakva rabota. > Mezhdu drugoto, sega useshtam che s tezi pool-ove testvaneto za leakove > njama da e tolkova prosto - kato cjalo leakove njama da ima, no mozhem da > leakvame na nivo pool, koeto pak e zle. > > Za collating sequences i equivallence classes (osobeno za purvoto), baja se > muchih i maj ne razbrah napulno kakva e idejata. Vse pak attachvam edin > html kojto gi specificira tija fenomeni. Sushto taka perllocale e dosta > polezen - tam sa objasneni s po-choveshki dumi. Drugoto e re.html, koeto e > v libfsd/doc -- tam e gramatikata na extended reguljaren izraz. > > On Friday 30 September 2005 12:51, Petar Petrov wrote: > > Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. Gledam, > > che si imame i lexer (rex_scanner.c). > > > > Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C code. > > Ako vzemem code-a ot at-framework-a shte trqbva da go dobavim v libfsd > > distroto, a tova shte e ~40% ot celiq code. Koeto e malko typo. > > > > Shte vidq dali shte moga da dobavq collating sequences i equivalence > > classes v reg ex. parser-a. > > > > On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > > Prashtam distroto dokudeto sum go dobutal. > > > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i > > > sushtevremenno lesno da se programirat i nakraja napisah edin super > > > generic > > > ciklichen single-linked list, podoben na NC-Base list-a, obache jade > > > po-malko > > > pamet i polzva pool za da si allocira node-ovete. Pool-a se podava i > > > globalen > > > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > > > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se alocira > > > pamet za > > > nego. Tova e za da se embedva v drugi strukturi, za koito ima pool > > > (state_t i > > > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s > > > malloc e samijat avtomat. > > > Pesho - rex parser-a - nadjavam se da se izkefish :) > > > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem > > > osnovite > > > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto da e > > > dostatuchno generic taka che da ne zavisim ot neshto, koeto trjabva da > > > se instalira (neshto koeto go njama vuv vsjaka moderna linux > > > distribucija). Variant e na perl test frameworka primerno. No trjabva > > > neshto da se izmisli > > > za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno > > > posle da > > > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama da go > > > napravim izobshto) a mi se iska vse pak da imame njakakuv nachin da si > > > podsigurim kachestvoto na lib-a inache ne si struva zagubenoto vreme. > > > Sushto variant e da si vzemem ot NC-Base at_framework-a i da go nagodim > > > (i bez > > > drugo nie sme si go mislili i sme sviknali da si go pozvame) > > > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto vidite > > > kato > > > problem -- kazvajte. > > > > > > Ajde leka i pishete. "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 10/03/05 00:34:11 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-09-30 10:08:49
|
Za autotest framework-a, misleh si che ako go dobavim v distroto shte go slozhim otdelno v libfsd/autotest foldera i njama nito da se instalira nito da se nabiva na ochi. ot druga strana bi ni spestilo razni neshta (memory leak test naprimer kakto i formatting-a na rezultatite. Demek sus sigurnost shte si spestim njakakva rabota. Mezhdu drugoto, sega useshtam che s tezi pool-ove testvaneto za leakove njama da e tolkova prosto - kato cjalo leakove njama da ima, no mozhem da leakvame na nivo pool, koeto pak e zle. Za collating sequences i equivallence classes (osobeno za purvoto), baja se muchih i maj ne razbrah napulno kakva e idejata. Vse pak attachvam edin html kojto gi specificira tija fenomeni. Sushto taka perllocale e dosta polezen - tam sa objasneni s po-choveshki dumi. Drugoto e re.html, koeto e v libfsd/doc -- tam e gramatikata na extended reguljaren izraz. On Friday 30 September 2005 12:51, Petar Petrov wrote: > Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. Gledam, che > si imame i lexer (rex_scanner.c). > > Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C code. Ako > vzemem code-a ot at-framework-a shte trqbva da go dobavim v libfsd > distroto, a tova shte e ~40% ot celiq code. Koeto e malko typo. > > Shte vidq dali shte moga da dobavq collating sequences i equivalence > classes v reg ex. parser-a. > > On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > Prashtam distroto dokudeto sum go dobutal. > > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i > > sushtevremenno lesno da se programirat i nakraja napisah edin super > > generic > > ciklichen single-linked list, podoben na NC-Base list-a, obache jade > > po-malko > > pamet i polzva pool za da si allocira node-ovete. Pool-a se podava i > > globalen > > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se alocira > > pamet za > > nego. Tova e za da se embedva v drugi strukturi, za koito ima pool > > (state_t i > > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s malloc > > e samijat avtomat. > > Pesho - rex parser-a - nadjavam se da se izkefish :) > > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem > > osnovite > > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto da e > > dostatuchno generic taka che da ne zavisim ot neshto, koeto trjabva da se > > instalira (neshto koeto go njama vuv vsjaka moderna linux distribucija). > > Variant e na perl test frameworka primerno. No trjabva neshto da se > > izmisli > > za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno posle > > da > > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama da go > > napravim izobshto) a mi se iska vse pak da imame njakakuv nachin da si > > podsigurim kachestvoto na lib-a inache ne si struva zagubenoto vreme. > > Sushto variant e da si vzemem ot NC-Base at_framework-a i da go nagodim > > (i bez > > drugo nie sme si go mislili i sme sviknali da si go pozvame) > > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto vidite > > kato > > problem -- kazvajte. > > > > Ajde leka i pishete. |
|
From: Petar P. <pes...@gm...> - 2005-09-30 09:51:42
|
Razgledah code-a, che go i razbrah :). Reg ex. parser-a kyrti. Gledam, che si imame i lexer (rex_scanner.c). Za AT-to, na men mi se struva naj - dobre da si gi pishem kato C code. Ako vzemem code-a ot at-framework-a shte trqbva da go dobavim v libfsd distroto= , a tova shte e ~40% ot celiq code. Koeto e malko typo. Shte vidq dali shte moga da dobavq collating sequences i equivalence classe= s v reg ex. parser-a. On 9/30/05, Ivan Peikov <va...@ne...> wrote: > > Prashtam distroto dokudeto sum go dobutal. > Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i > sushtevremenno lesno da se programirat i nakraja napisah edin super > generic > ciklichen single-linked list, podoben na NC-Base list-a, obache jade > po-malko > pamet i polzva pool za da si allocira node-ovete. Pool-a se podava i > globalen > v ramkite na avtomata. Tova reshava problema s fragmentacijata. > Oburnete vnimanie sushto che fsd_list_t se init-va bez da se alocira pame= t > za > nego. Tova e za da se embedva v drugi strukturi, za koito ima pool > (state_t i > trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s malloc = e > samijat avtomat. > Pesho - rex parser-a - nadjavam se da se izkefish :) > Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem > osnovite > mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto da e > dostatuchno generic taka che da ne zavisim ot neshto, koeto trjabva da se > instalira (neshto koeto go njama vuv vsjaka moderna linux distribucija). > Variant e na perl test frameworka primerno. No trjabva neshto da se > izmisli > za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno posle > da > pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama da go > napravim izobshto) a mi se iska vse pak da imame njakakuv nachin da si > podsigurim kachestvoto na lib-a inache ne si struva zagubenoto vreme. > Sushto variant e da si vzemem ot NC-Base at_framework-a i da go nagodim (= i > bez > drugo nie sme si go mislili i sme sviknali da si go pozvame) > Povecheto neshta izobshto ne sa testvani taka che vsichko koeto vidite > kato > problem -- kazvajte. > > Ajde leka i pishete. > > > > |
|
From: Ivan P. <va...@ne...> - 2005-09-30 07:35:57
|
Prashtam distroto dokudeto sum go dobutal. Sreshtnah baja problemi s listovete, kak da sa dostatuchno obshti i sushtevremenno lesno da se programirat i nakraja napisah edin super generic ciklichen single-linked list, podoben na NC-Base list-a, obache jade po-malko pamet i polzva pool za da si allocira node-ovete. Pool-a se podava i globalen v ramkite na avtomata. Tova reshava problema s fragmentacijata. Oburnete vnimanie sushto che fsd_list_t se init-va bez da se alocira pamet za nego. Tova e za da se embedva v drugi strukturi, za koito ima pool (state_t i trans_t). Po tozi nachin edinstvenoto koeto se alokira direktno s malloc e samijat avtomat. Pesho - rex parser-a - nadjavam se da se izkefish :) Ima i njakakvo nachalno TODO (v distroto). Za momenta, dokato kovem osnovite mi se struva naj-vazhno da izmislim nachin za autotestvane, kojto da e dostatuchno generic taka che da ne zavisim ot neshto, koeto trjabva da se instalira (neshto koeto go njama vuv vsjaka moderna linux distribucija). Variant e na perl test frameworka primerno. No trjabva neshto da se izmisli za da si pishem testovete zaedno s koda. Inache shte e mnogo trudno posle da pishem nakup (i tuj kato nikoj ne ni pritiska naj-verojatno njama da go napravim izobshto) a mi se iska vse pak da imame njakakuv nachin da si podsigurim kachestvoto na lib-a inache ne si struva zagubenoto vreme. Sushto variant e da si vzemem ot NC-Base at_framework-a i da go nagodim (i bez drugo nie sme si go mislili i sme sviknali da si go pozvame) Povecheto neshta izobshto ne sa testvani taka che vsichko koeto vidite kato problem -- kazvajte. Ajde leka i pishete. |
|
From: Petar P. <pes...@gm...> - 2005-09-26 14:47:43
|
ok, v 7 na halite. Tykmo razbrah, che imam 5.40 na dyrjavniq. Ne e otlichno no pyk nali sym barlavarin veche :), shte ima cherpalo (a i novata rabota syshto e povod). On 9/26/05, Georgi Mungov <geo...@tu...> wrote: > > ok > On Monday 26 September 2005 10:16, Ivan Peikov wrote: > > utre (vtornik) 7 na halite? > > > > On Monday 26 September 2005 10:17, Georgi Mungov wrote: > > > lib...@li... > > > > ------------------------------------------------------- > > SF.Net email is sponsored by: > > Tame your development challenges with Apache's Geronimo App Server. > > Download it for free - -and be entered to win a 42" plasma tv or your > very > > own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.ph= p > > _______________________________________________ > > Libfsd-devel mailing list > > Lib...@li... > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > > "Tumbleweed E-mail Firewall <tumbleweed.com <http://tumbleweed.com>>" mad= e > the following > annotations on 09/26/05 00:32:41 > > -------------------------------------------------------------------------= ----- > This e-mail, including attachments, may include confidential and/or > proprietary information, and may be used only by the person or entity to > which it is addressed. If the reader of this e-mail is not the intended > recipient or his or her authorized agent, the reader is hereby notified t= hat > any dissemination, distribution or copying of this e-mail is prohibited. = If > you have received this e-mail in error, please notify the sender by reply= ing > to this message and delete this e-mail immediately. > > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D > > > > ------------------------------------------------------- > SF.Net email is sponsored by: > Tame your development challenges with Apache's Geronimo App Server. > Download > it for free - -and be entered to win a 42" plasma tv or your very own > Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel > |
|
From: Georgi M. <geo...@tu...> - 2005-09-26 07:24:32
|
ok On Monday 26 September 2005 10:16, Ivan Peikov wrote: > utre (vtornik) 7 na halite? > > On Monday 26 September 2005 10:17, Georgi Mungov wrote: > > lib...@li... > > ------------------------------------------------------- > SF.Net email is sponsored by: > Tame your development challenges with Apache's Geronimo App Server. > Download it for free - -and be entered to win a 42" plasma tv or your very > own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php > _______________________________________________ > Libfsd-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libfsd-devel "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following annotations on 09/26/05 00:32:41 ------------------------------------------------------------------------------ This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately. ============================================================================== |
|
From: Ivan P. <va...@ne...> - 2005-09-26 07:20:13
|
utre (vtornik) 7 na halite? On Monday 26 September 2005 10:17, Georgi Mungov wrote: > lib...@li... |
|
From: Georgi M. <geo...@tu...> - 2005-09-26 07:18:05
|
vtornik, sriada?
On Wednesday 21 September 2005 14:09, Ivan Peikov wrote:
> Den, chas i mjasto? Az sum navit za proizvolen den drugata sedmica, samo da
> e sled 6.
> Pesho -- kakvo stana s durzhavnija? Ot teb li shte sa birite?
>
> On Wednesday 21 September 2005 13:32, Georgi Mungov wrote:
> > edna bira drugata sedmica
> >
> > On Monday 19 September 2005 15:52, Ivan Peikov wrote:
> > > ami ne che ne mozhe, obache povecheto algoritmi za koito se seshtam sa
> > > ot vida: "ami stignah do njakakuv state i sega za vsichki states, do
> > > koito imam transition ot tozi state napravi edi kakvo si". i za tova mi
> > > se vidja naj-udachno (razbiraj efektivno) da imame direkten dostup do
> > > transitionite ot vseki state
> > >
> > > On Monday 19 September 2005 15:38, Georgi Mungov wrote:
> > > > ami moje bi vmesto list transitions v state-a da sa v multimap ?
> > > >
> > > > On Saturday 17 September 2005 10:37, Ivan Peikov wrote:
> > > > > Privet!
> > > > >
> > > > > Za da ne zamira "proekta" (:]) dokato Pesho go trese predpizpitna
> > > > > treska (uspeh, Pesho!), eto kakvo e polozhenieto i kakvi vuprosi me
> > > > > zanimavat.
> > > > >
> > > > > Znachi do tuk imam implementnat parser-a na reguljarni izrazi,
> > > > > kojto e pochti POSIX-compatible (ostavat mnogo malko featuri da se
> > > > > dobavjat, ama za sega me murzi, puk i te sa njakvi dosta nenuzhni -
> > > > > nikoga ne bjah chuval za collating sequences i equivallence classes
> > > > > v reguljarni izrazi predi da procheta spesifikacijata). Taka che za
> > > > > momenta tozi parser shte svurshi rabota. Edinstveno ostavat da se
> > > > > napravjat testove i malko da se pozagladjat njakoi rubove :).
> > > > >
> > > > > Mislja che ottuk mozhem da zapochnem da pravim avtomatite. Struva
> > > > > mi se naj-dobre da zapochnem s njakakuv vid general avtomati (GFA -
> > > > > General Finite Automaton), kojto da e abstrakten i da njama azbuka.
> > > > > Tuk idva purvijat vupros. Izglezhda edinstvenoto, koeto razlichava
> > > > > vsichki vidove avtomati i transduceri za koito se seshtam e
> > > > > azbukata. Primerno za NFA tova shte sa njakakvi simvoli, a za
> > > > > transducerite azbukata sa dvojki ot naj-obshto simvol i output
> > > > > string. Toest trjabva tozi GFA da se napravi taka che sled tova da
> > > > > mozhe s neshto kato konfiguracija da se napravi proizvolen drug
> > > > > avtomat. Za momenta varianta, do kojto sum go dobutal, e da ima
> > > > > slednite obekti
> > > > >
> > > > >
> > > > > *************************************************************
> > > > >
> > > > > struct trans;
> > > > > struct state;
> > > > >
> > > > > struct gfa {
> > > > > struct state *states; // Head of the list of states
> > > > > struct state *start; // Head of the list of starting states
> > > > > struct state *final; // Head of the list of final states
> > > > > pool_t state_pool; // Pool from which the states are allocated
> > > > > pool_t trans_pool; // Pool from which the transitions are
> > > > > allocated size_t state_nr; // Number of states - used for
> > > > > assigning next state id // --- Some overridable methods for
> > > > > managing transitions ----------------- insert_func_t *insert; //
> > > > > Transition inserting function remove_func_t *remove; // Transition
> > > > > removing function search_func_t *search; // Transition searching
> > > > > function
> > > > >
> > > > > };
> > > > >
> > > > > struct state {
> > > > > struct trans *incoming; // List of incoming transitions
> > > > > struct trans *outgoing; // List of outgoing transitions
> > > > > struct state *start; // Link in list of start states (NULL if
> > > > > non-start) struct state *final; // Link in list of final states
> > > > > (NULL if non-final) struct state *next; // Link in list of all
> > > > > states struct gfa *parent; // The automaton this state is part of
> > > > > size_t id; // Id for dumping and accessing purposes
> > > > > // More stuff?...
> > > > > };
> > > > >
> > > > > struct trans {
> > > > > struct state *src; // Where from
> > > > > struct state *dst; // Where to
> > > > > struct trans *src_next; // Link in the list of transitions for
> > > > > src struct trans *dst_next; // Link in the list of transitions for
> > > > > dst // Intentionally no symbol -- this is an abstract transition };
> > > > >
> > > > > // --- Example: NFA transition
> > > > >
> > > > > struct nfa_trans {
> > > > > struct trans base; // Should be first for lists and casts to work
> > > > > char alpha;
> > > > > };
> > > > >
> > > > >
> > > > > *************************************************************
> > > > >
> > > > >
> > > > >
> > > > > Neshto ej takova.. Puk posle kato se initializira konkreten avtomat
> > > > > shte se podava size-a na alphabet symbola i taka shte se
> > > > > inicializira trans_pool. Sushto shte se podavat i konkretni funkcii
> > > > > za
> > > > > insert/remove/search na transition, koito obshto vzeto shte vzimat
> > > > > src_state, symbol, dst_state i shte pravjat kakvoto tam se nalaga.
> > > > >
> > > > > Tuk veche idva vuprosa za efektivnostta. Mozhe da vi pravi
> > > > > vpechatlenie che baja pamet shte jade tova zhivotno, obache puk
> > > > > shte ima vsichki neobhodimi vruzki, koeto se nadjavam da ulesni
> > > > > algoritmite posle.
> > > > >
> > > > > Stoi sushto vuprosa kak shte sa podredeni transitionite za daden
> > > > > state. Za momenta shte e dummy list bez nikakva naredba, obache se
> > > > > seshtate che tova ni kachva slozhnostta na izpulnenie na avtomat
> > > > > vurhu dadena duma. Tova e edna ot prichinite da zabija
> > > > > insert/remove/search operaciite kato function pointers - edin vid
> > > > > vseki konkreten avtomat shte si reshava kak da gi redi, i tursi.
> > > > > Mozhe za momenta da gi ostavim po tozi nachin puk v budeshte ako se
> > > > > poluchi mnogo tezhko naistina da mislim kak da se optimizira.
> > > > > Obshto vzeto ima megdan.
> > > > >
> > > > >
> > > > > Ta tova e obshto vzeto predlozhenieto za sega. Kakvo mislite?
> > > > > Mozhe da obsudim na zhivo Pesho kato si vzeme durzhavnija s
> > > > > otlichie, i bez tova shte trjabva da cherpi ne edna i dve biri po
> > > > > povoda :)
> > > > >
> > > > >
> > > > > Ajde do kukane i pishete ako neshto
> > > > >
> > > > >
> > > > >
> > > > > -------------------------------------------------------
> > > > > SF.Net email is sponsored by:
> > > > > Tame your development challenges with Apache's Geronimo App Server.
> > > > > Download it for free - -and be entered to win a 42" plasma tv or
> > > > > your very own Sony(tm)PSP. Click here to play:
> > > > > http://sourceforge.net/geronimo.php
> > > > > _______________________________________________
> > > > > Libfsd-devel mailing list
> > > > > Lib...@li...
> > > > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
> > > >
> > > > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following
> > > > annotations on 09/19/05 05:46:51
> > > > ---------------------------------------------------------------------
> > > >-- -- -- --- This e-mail, including attachments, may include
> > > > confidential and/or proprietary information, and may be used only by
> > > > the person or entity to which it is addressed. If the reader of this
> > > > e-mail is not the intended recipient or his or her authorized agent,
> > > > the reader is hereby notified that any dissemination, distribution or
> > > > copying of this e-mail is prohibited. If you have received this
> > > > e-mail in error, please notify the sender by replying to this message
> > > > and delete this e-mail immediately.
> > > > =====================================================================
> > > >== == == ===
> > > >
> > > >
> > > >
> > > > -------------------------------------------------------
> > > > SF.Net email is sponsored by:
> > > > Tame your development challenges with Apache's Geronimo App Server.
> > > > Download it for free - -and be entered to win a 42" plasma tv or your
> > > > very own Sony(tm)PSP. Click here to play:
> > > > http://sourceforge.net/geronimo.php
> > > > _______________________________________________
> > > > Libfsd-devel mailing list
> > > > Lib...@li...
> > > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
> > >
> > > -------------------------------------------------------
> > > SF.Net email is sponsored by:
> > > Tame your development challenges with Apache's Geronimo App Server.
> > > Download it for free - -and be entered to win a 42" plasma tv or your
> > > very own Sony(tm)PSP. Click here to play:
> > > http://sourceforge.net/geronimo.php
> > > _______________________________________________
> > > Libfsd-devel mailing list
> > > Lib...@li...
> > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
> >
> > "Tumbleweed E-mail Firewall <tumbleweed.com>" made the following
> > annotations on 09/21/05 03:40:44
> > -------------------------------------------------------------------------
> >-- --- This e-mail, including attachments, may include confidential and/or
> > proprietary information, and may be used only by the person or entity to
> > which it is addressed. If the reader of this e-mail is not the intended
> > recipient or his or her authorized agent, the reader is hereby notified
> > that any dissemination, distribution or copying of this e-mail is
> > prohibited. If you have received this e-mail in error, please notify the
> > sender by replying to this message and delete this e-mail immediately.
> > =========================================================================
> >== ===
> >
> >
> >
> > -------------------------------------------------------
> > SF.Net email is sponsored by:
> > Tame your development challenges with Apache's Geronimo App Server.
> > Download it for free - -and be entered to win a 42" plasma tv or your
> > very own Sony(tm)PSP. Click here to play:
> > http://sourceforge.net/geronimo.php
> > _______________________________________________
> > Libfsd-devel mailing list
> > Lib...@li...
> > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
>
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server.
> Download it for free - -and be entered to win a 42" plasma tv or your very
> own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Libfsd-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libfsd-devel
"Tumbleweed E-mail Firewall <tumbleweed.com>" made the following
annotations on 09/26/05 00:26:17
------------------------------------------------------------------------------
This e-mail, including attachments, may include confidential and/or proprietary information, and may be used only by the person or entity to which it is addressed. If the reader of this e-mail is not the intended recipient or his or her authorized agent, the reader is hereby notified that any dissemination, distribution or copying of this e-mail is prohibited. If you have received this e-mail in error, please notify the sender by replying to this message and delete this e-mail immediately.
==============================================================================
|
|
From: Petar P. <pes...@gm...> - 2005-09-21 11:16:07
|
Moje, dyrjavniq mina. Sega chakame sys zataen dyh da vidik k'vo shte izleze=
=20
:)).
On 9/21/05, Georgi Mungov <geo...@tu...> wrote:
>=20
> edna bira drugata sedmica
>=20
> On Monday 19 September 2005 15:52, Ivan Peikov wrote:
> > ami ne che ne mozhe, obache povecheto algoritmi za koito se seshtam sa=
=20
> ot
> > vida: "ami stignah do njakakuv state i sega za vsichki states, do koito
> > imam transition ot tozi state napravi edi kakvo si". i za tova mi se=20
> vidja
> > naj-udachno (razbiraj efektivno) da imame direkten dostup do=20
> transitionite
> > ot vseki state
> >
> > On Monday 19 September 2005 15:38, Georgi Mungov wrote:
> > > ami moje bi vmesto list transitions v state-a da sa v multimap ?
> > >
> > > On Saturday 17 September 2005 10:37, Ivan Peikov wrote:
> > > > Privet!
> > > >
> > > > Za da ne zamira "proekta" (:]) dokato Pesho go trese predpizpitna
> > > > treska (uspeh, Pesho!), eto kakvo e polozhenieto i kakvi vuprosi me
> > > > zanimavat.
> > > >
> > > > Znachi do tuk imam implementnat parser-a na reguljarni izrazi, kojt=
o=20
> e
> > > > pochti POSIX-compatible (ostavat mnogo malko featuri da se dobavjat=
,
> > > > ama za sega me murzi, puk i te sa njakvi dosta nenuzhni - nikoga ne
> > > > bjah chuval za collating sequences i equivallence classes v=20
> reguljarni
> > > > izrazi predi da procheta spesifikacijata). Taka che za momenta tozi
> > > > parser shte svurshi rabota. Edinstveno ostavat da se napravjat=20
> testove
> > > > i malko da se pozagladjat njakoi rubove :).
> > > >
> > > > Mislja che ottuk mozhem da zapochnem da pravim avtomatite. Struva m=
i=20
> se
> > > > naj-dobre da zapochnem s njakakuv vid general avtomati (GFA -=20
> General
> > > > Finite Automaton), kojto da e abstrakten i da njama azbuka. Tuk idv=
a
> > > > purvijat vupros. Izglezhda edinstvenoto, koeto razlichava vsichki
> > > > vidove avtomati i transduceri za koito se seshtam e azbukata.=20
> Primerno
> > > > za NFA tova shte sa njakakvi simvoli, a za transducerite azbukata s=
a
> > > > dvojki ot naj-obshto simvol i output string. Toest trjabva tozi GFA=
=20
> da
> > > > se napravi taka che sled tova da mozhe s neshto kato konfiguracija=
=20
> da
> > > > se napravi proizvolen drug avtomat. Za momenta varianta, do kojto=
=20
> sum
> > > > go dobutal, e da ima slednite obekti
> > > >
> > > >
> > > > *************************************************************
> > > >
> > > > struct trans;
> > > > struct state;
> > > >
> > > > struct gfa {
> > > > struct state *states; // Head of the list of states
> > > > struct state *start; // Head of the list of starting states
> > > > struct state *final; // Head of the list of final states
> > > > pool_t state_pool; // Pool from which the states are allocated
> > > > pool_t trans_pool; // Pool from which the transitions are allocated
> > > > size_t state_nr; // Number of states - used for assigning next
> > > > state id // --- Some overridable methods for managing transitions
> > > > ----------------- insert_func_t *insert; // Transition inserting
> > > > function remove_func_t *remove; // Transition removing function
> > > > search_func_t *search; // Transition searching function
> > > >
> > > > };
> > > >
> > > > struct state {
> > > > struct trans *incoming; // List of incoming transitions
> > > > struct trans *outgoing; // List of outgoing transitions
> > > > struct state *start; // Link in list of start states (NULL if
> > > > non-start) struct state *final; // Link in list of final states=20
> (NULL
> > > > if non-final) struct state *next; // Link in list of all states
> > > > struct gfa *parent; // The automaton this state is part of
> > > > size_t id; // Id for dumping and accessing purposes
> > > > // More stuff?...
> > > > };
> > > >
> > > > struct trans {
> > > > struct state *src; // Where from
> > > > struct state *dst; // Where to
> > > > struct trans *src_next; // Link in the list of transitions for src
> > > > struct trans *dst_next; // Link in the list of transitions for dst
> > > > // Intentionally no symbol -- this is an abstract transition
> > > > };
> > > >
> > > > // --- Example: NFA transition
> > > >
> > > > struct nfa_trans {
> > > > struct trans base; // Should be first for lists and casts to work
> > > > char alpha;
> > > > };
> > > >
> > > >
> > > > *************************************************************
> > > >
> > > >
> > > >
> > > > Neshto ej takova.. Puk posle kato se initializira konkreten avtomat
> > > > shte se podava size-a na alphabet symbola i taka shte se=20
> inicializira
> > > > trans_pool. Sushto shte se podavat i konkretni funkcii za
> > > > insert/remove/search na transition, koito obshto vzeto shte vzimat
> > > > src_state, symbol, dst_state i shte pravjat kakvoto tam se nalaga.
> > > >
> > > > Tuk veche idva vuprosa za efektivnostta. Mozhe da vi pravi=20
> vpechatlenie
> > > > che baja pamet shte jade tova zhivotno, obache puk shte ima vsichki
> > > > neobhodimi vruzki, koeto se nadjavam da ulesni algoritmite posle.
> > > >
> > > > Stoi sushto vuprosa kak shte sa podredeni transitionite za daden=20
> state.
> > > > Za momenta shte e dummy list bez nikakva naredba, obache se seshtat=
e
> > > > che tova ni kachva slozhnostta na izpulnenie na avtomat vurhu daden=
a
> > > > duma. Tova e edna ot prichinite da zabija insert/remove/search
> > > > operaciite kato function pointers - edin vid vseki konkreten avtoma=
t
> > > > shte si reshava kak da gi redi, i tursi. Mozhe za momenta da gi=20
> ostavim
> > > > po tozi nachin puk v budeshte ako se poluchi mnogo tezhko naistina=
=20
> da
> > > > mislim kak da se optimizira. Obshto vzeto ima megdan.
> > > >
> > > >
> > > > Ta tova e obshto vzeto predlozhenieto za sega. Kakvo mislite?
> > > > Mozhe da obsudim na zhivo Pesho kato si vzeme durzhavnija s=20
> otlichie, i
> > > > bez tova shte trjabva da cherpi ne edna i dve biri po povoda :)
> > > >
> > > >
> > > > Ajde do kukane i pishete ako neshto
> > > >
> > > >
> > > >
> > > > -------------------------------------------------------
> > > > SF.Net email is sponsored by:
> > > > Tame your development challenges with Apache's Geronimo App Server.
> > > > Download it for free - -and be entered to win a 42" plasma tv or=20
> your
> > > > very own Sony(tm)PSP. Click here to play:
> > > > http://sourceforge.net/geronimo.php
> > > > _______________________________________________
> > > > Libfsd-devel mailing list
> > > > Lib...@li...
> > > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
> > >
> > > "Tumbleweed E-mail Firewall <tumbleweed.com <http://tumbleweed.com>>"=
=20
> made the following
> > > annotations on 09/19/05 05:46:51
> > >=20
> -------------------------------------------------------------------------
> > >-- --- This e-mail, including attachments, may include confidential=20
> and/or
> > > proprietary information, and may be used only by the person or entity=
=20
> to
> > > which it is addressed. If the reader of this e-mail is not the=20
> intended
> > > recipient or his or her authorized agent, the reader is hereby=20
> notified
> > > that any dissemination, distribution or copying of this e-mail is
> > > prohibited. If you have received this e-mail in error, please notify=
=20
> the
> > > sender by replying to this message and delete this e-mail immediately=
.
> > >=20
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> > >=3D=3D =3D=3D=3D
> > >
> > >
> > >
> > > -------------------------------------------------------
> > > SF.Net email is sponsored by:
> > > Tame your development challenges with Apache's Geronimo App Server.
> > > Download it for free - -and be entered to win a 42" plasma tv or your
> > > very own Sony(tm)PSP. Click here to play:
> > > http://sourceforge.net/geronimo.php
> > > _______________________________________________
> > > Libfsd-devel mailing list
> > > Lib...@li...
> > > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
> >
> > -------------------------------------------------------
> > SF.Net email is sponsored by:
> > Tame your development challenges with Apache's Geronimo App Server.
> > Download it for free - -and be entered to win a 42" plasma tv or your=
=20
> very
> > own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.ph=
p
> > _______________________________________________
> > Libfsd-devel mailing list
> > Lib...@li...
> > https://lists.sourceforge.net/lists/listinfo/libfsd-devel
>=20
> "Tumbleweed E-mail Firewall <tumbleweed.com <http://tumbleweed.com>>" mad=
e=20
> the following
> annotations on 09/21/05 03:40:44
>=20
> -------------------------------------------------------------------------=
-----
> This e-mail, including attachments, may include confidential and/or=20
> proprietary information, and may be used only by the person or entity to=
=20
> which it is addressed. If the reader of this e-mail is not the intended=
=20
> recipient or his or her authorized agent, the reader is hereby notified t=
hat=20
> any dissemination, distribution or copying of this e-mail is prohibited. =
If=20
> you have received this e-mail in error, please notify the sender by reply=
ing=20
> to this message and delete this e-mail immediately.
>=20
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D
>=20
>=20
>=20
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server.=20
> Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Libfsd-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libfsd-devel
>
|