I-DuckDB yinkqubo yolawulo lwedatha ye-SQL OLAP ekwenzeni
Ukukhutshwa kwenguqulelo entsha ye-DBMS kwabhengezwa I-DuckDB 0.6.0, uguqulelo apho Uxinzelelo lwedatha luphuculwe, ukongeza kwinto yokuba imisebenzi emitsha yongezwe, kunye nokuphuculwa kokugcinwa, phakathi kwezinye izinto.
I-DuckDB dibanisa iipropati zeSQLite njengokubumbana, ukukwazi ukudibanisa ngendlela yelayibrari edibeneyo, ukugcinwa kwedatha kwifayile enye kunye ne-interface ye-CLI efanelekileyo, kunye nezixhobo kunye nokulungiswa kokwenza imibuzo yohlalutyo egubungela inxalenye ebalulekileyo yedatha egciniweyo, umzekelo, eyenza udityaniso lwemixholo yetafile okanye ukudibanisa iitafile ezininzi ezinkulu.
Iimpawu eziphambili ezintsha zeDuckDB 0.6.0
Kolu guqulelo lutsha luvezwayo, kugxininiswe oko umsebenzi waqhubeka ekuphuculeni ifomathi yokugcina, ngaphandle koko indlela yokubhala idisk iphunyeziwe, apho xa isethi enkulu yedatha ilayishwe kwi-transaction enye, idatha ixinzelelwe kwaye ihanjiswe kwifayile esuka kwisiseko sedatha ngaphandle kokulinda umyalelo we-COMMIT ukwenza ukuthengiselana.
Olunye utshintsho olugqamayo kwinguqulelo entsha kukuba inkxaso eyongeziweyo yokulayisha ngokuhambelanayo kwedatha kwiitheyibhile ezahlukeneyo, enokunyusa kakhulu isantya sokulayisha kwiinkqubo ezininzi. Ngokomzekelo, kwinguqulo yakudala, ukulayisha i-database kunye ne-150 yezigidi zemiqolo kwi-CPU ye-10-core ithathe imizuzwana ye-91, kwaye kwinguqu entsha, lo msebenzi uthatha imizuzwana eyi-17. Kukho iindlela ezimbini zokulayisha okufanayo: ngokugcinwa komyalelo werekhodi kwaye ngaphandle kokugcinwa komyalelo.
Ukunyanzeliswa kwedatha, i-algorithm ye-FSST isetyenziswa (Iimpawu ezimileyo zeTheyibhile eKhawulezayo), ekuvumela ukuba upakishe idatha ngaphakathi kwemigca usebenzisa isichazi-magama esiqhelekileyo sohlobo lwematshisi. Ukusetyenziswa kwe-algorithm entsha kuvunyelwe ukunciphisa ubungakanani bedatha yovavanyo ukusuka kwi-761 MB ukuya kwi-251 MB.
Ukucinezela amanani (DOUBLE kunye neFLOAT) i-Chimp kunye ne-Patas algorithms zicetywayo. Xa kuthelekiswa ne-algorithm yeGorilla yangaphambili, iChimp ibonelela ngenqanaba eliphezulu loxinzelelo kunye noxinzelelo olukhawulezayo. I-algorithm ye-Patas isala ngasemva kweChimp ngokwemiqathango yoxinzelelo, kodwa ikhawuleza kakhulu kwisantya soxinzelelo, esifana nqwa nokufunda idatha engaxinzelelwanga.
Kukwaphawulwe ukuba yongezwa isakhono sokulinga ukulayisha idatha kwiifayile ze-CSV kwimijelo emininzi ehambelanayo (SET experimental_parallel_csv=true), enciphisa kakhulu ixesha lokulayisha kwiifayile ezinkulu ze-CSV. Umzekelo, xa ukhetho lwenziwe, ixesha lokukhuphela lefayile ye-CSV ye-720MB lancitshiswa ukusuka kwimizuzwana eyi-3,5 ukuya kwimizuzwana eyi-0,6.
Olunye utshintsho evelele kule nguqulo intsha:
- Ukuphunyezwa okufanayo kokudalwa kwesalathisi kunye nemisebenzi yokulawula iphunyeziwe.
- I-SQL ibonelela ngokukwazi ukwenza imibuzo eqala ngegama elithi "UKUSUKA" endaweni yokuthi "KHETHA". Kulo mzekelo, umbuzo ucingelwa ukuqala ngo "KHETHA *".
- Inkxaso eyongeziweyo ye "COLUMNS" intetho kwiSQL, ikuvumela ukuba wenze umsebenzi kwimihlathi emininzi ngaphandle kokuphinda-phinda ibinzana.
- Ukusetyenziswa kwememori okulungiselelwe. Ngokungagqibekanga kwiqonga leLinux, ilayibrari yejemalloc isetyenziselwa ulawulo lwenkumbulo. Kuphuculwe kakhulu ukusebenza kodibaniso lwe-hash xa inkumbulo ilinganiselwe.
- Yongezwe ".imode ye-duckbox" imo yemveliso kwi-CLI, ukulahlwa kweentsika eziziko ngokusekelwe kububanzi bemigca yefestile ye-terminal). Nge ".maxrows X" ipharamitha, unganciphisa inani lemiqolo yemveliso.
- I-CLI ibonelela ngegalelo lokuzigqibezela ngokuzenzekelayo (amagama angundoqo, amagama etafile, imisebenzi, amagama eekholamu, kunye nokufakwa kwamagama eefayile kugqityiwe).
- I-CLI yenziwe ngokungagqibekanga ukubonisa isalathi senkqubela yombuzo.
ekugqibeleni ukuba ukhona unomdla wokwazi ngakumbi ngayo, ungajonga iinkcukacha Kule khonkco ilandelayo.