{"id":32693,"date":"2025-10-24T22:32:51","date_gmt":"2025-10-24T19:32:51","guid":{"rendered":"https:\/\/fatihsoysal.com\/blog\/python-ile-yaris-durumlarini-yenmek-coklu-is-parcaciginda-kaosu-evcillestirmek\/"},"modified":"2025-10-24T22:32:51","modified_gmt":"2025-10-24T19:32:51","slug":"python-ile-yaris-durumlarini-yenmek-coklu-is-parcaciginda-kaosu-evcillestirmek","status":"publish","type":"post","link":"https:\/\/fatihsoysal.com\/blog\/python-ile-yaris-durumlarini-yenmek-coklu-is-parcaciginda-kaosu-evcillestirmek\/","title":{"rendered":"Python ile Yar\u0131\u015f Durumlar\u0131n\u0131 Yenmek: \u00c7oklu \u0130\u015f Par\u00e7ac\u0131\u011f\u0131nda Kaosu Evcille\u015ftirmek"},"content":{"rendered":"

<\/p>\n

Modern yaz\u0131l\u0131m geli\u015ftirmenin ayr\u0131lmaz bir par\u00e7as\u0131 olan \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 (multithreading), uygulamalar\u0131m\u0131za e\u015f zamanl\u0131l\u0131k yetene\u011fi kazand\u0131rarak performans ve kullan\u0131c\u0131 deneyimi a\u00e7\u0131s\u0131ndan \u00f6nemli avantajlar sunar. Ancak bu g\u00fc\u00e7l\u00fc ara\u00e7, yanl\u0131\u015f kullan\u0131ld\u0131\u011f\u0131nda “yar\u0131\u015f durumlar\u0131” (race conditions) gibi \u00f6ng\u00f6r\u00fclemeyen ve hata ay\u0131klamas\u0131 zor sorunlara yol a\u00e7abilir. Bu makalede, Python d\u00fcnyas\u0131nda bu t\u00fcr kaoslar\u0131 nas\u0131l evcille\u015ftirece\u011fimizi, yar\u0131\u015f durumlar\u0131n\u0131n ne oldu\u011funu, neden ortaya \u00e7\u0131kt\u0131\u011f\u0131n\u0131 ve bunlar\u0131 \u00f6nlemek i\u00e7in hangi etkili senkronizasyon mekanizmalar\u0131n\u0131 kullanabilece\u011fimizi ad\u0131m ad\u0131m inceleyece\u011fiz. E\u011fer payla\u015f\u0131ml\u0131 kaynaklara eri\u015fimde ya\u015fanan belirsizliklerden b\u0131kt\u0131ysan\u0131z veya daha sa\u011flam, hata toleransl\u0131 \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalar geli\u015ftirmek istiyorsan\u0131z, do\u011fru yerdesiniz.<\/p>\n

\u00c7oklu i\u015f par\u00e7ac\u0131\u011f\u0131, bir program\u0131n ayn\u0131 anda birden fazla g\u00f6rev y\u00fcr\u00fctmesine olanak tan\u0131yan bir programlama modelidir. Genellikle, CPU’nun bir g\u00f6revi ask\u0131ya al\u0131p di\u011ferine ge\u00e7mesiyle “e\u015f zamanl\u0131” (concurrent) bir y\u00fcr\u00fctme hissi yarat\u0131l\u0131r, ancak modern \u00e7ok \u00e7ekirdekli i\u015flemcilerde bu g\u00f6revler ger\u00e7ekten “paralel” (parallel) olarak da \u00e7al\u0131\u015fabilirler. Python’da threading<\/code> mod\u00fcl\u00fc, bu yetene\u011fi bize sunar. Diyelim ki bir web uygulaman\u0131z var ve ayn\u0131 anda hem kullan\u0131c\u0131 iste\u011fini i\u015flemek hem de arka planda bir veritaban\u0131 temizleme i\u015flemi yapmak istiyorsunuz. \u0130\u015fte bu noktada \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 devreye girer. Her bir g\u00f6rev ayr\u0131 bir i\u015f par\u00e7ac\u0131\u011f\u0131na atan\u0131r ve i\u015fletim sistemi taraf\u0131ndan bu i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda h\u0131zl\u0131ca ge\u00e7i\u015f yap\u0131larak t\u00fcm g\u00f6revlerin e\u015f zamanl\u0131 olarak ilerledi\u011fi izlenimi verilir. Ancak bu e\u015f zamanl\u0131 y\u00fcr\u00fctme, ayn\u0131 bellekteki payla\u015f\u0131ml\u0131 kaynaklara (\u00f6rne\u011fin, bir de\u011fi\u015fken, bir liste veya bir dosya) birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 ayn\u0131 anda eri\u015fmeye \u00e7al\u0131\u015ft\u0131\u011f\u0131nda sorunlara yol a\u00e7abilir.<\/p>\n

Yar\u0131\u015f durumu (race condition), birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n ayn\u0131 payla\u015f\u0131ml\u0131 kayna\u011fa e\u015f zamanl\u0131 olarak eri\u015fmeye \u00e7al\u0131\u015ft\u0131\u011f\u0131nda ve bu eri\u015fimlerin s\u0131ras\u0131n\u0131n \u00f6ng\u00f6r\u00fclemez olmas\u0131ndan dolay\u0131 program\u0131n davran\u0131\u015f\u0131n\u0131n da \u00f6ng\u00f6r\u00fclemez hale gelmesidir. K\u0131sacas\u0131, bir i\u015f par\u00e7ac\u0131\u011f\u0131 bir kayna\u011f\u0131 de\u011fi\u015ftirirken, ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131 o kayna\u011f\u0131 okumaya veya de\u011fi\u015ftirmeye \u00e7al\u0131\u015f\u0131r ve bu durum tutars\u0131z veya hatal\u0131 sonu\u00e7lara yol a\u00e7ar. \u00d6rne\u011fin, iki i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n ayn\u0131 anda bir sayac\u0131 art\u0131rd\u0131\u011f\u0131n\u0131 d\u00fc\u015f\u00fcn\u00fcn. \u0130\u015f par\u00e7ac\u0131\u011f\u0131 A sayac\u0131n de\u011ferini okur, bir art\u0131r\u0131r. Tam bu s\u0131rada, \u0130\u015f par\u00e7ac\u0131\u011f\u0131 B de sayac\u0131n ayn\u0131 eski de\u011ferini okur, bir art\u0131r\u0131r. Her ikisi de ayn\u0131 anda yeni de\u011feri yazar. Sonu\u00e7 olarak, saya\u00e7 beklenen de\u011ferden daha az artm\u0131\u015f olur. Bu t\u00fcr hatalar, genellikle nadiren ortaya \u00e7\u0131kt\u0131klar\u0131 ve belirli y\u00fcr\u00fctme zamanlamalar\u0131na ba\u011fl\u0131 olduklar\u0131 i\u00e7in hata ay\u0131klamas\u0131 olduk\u00e7a g\u00fc\u00e7t\u00fcr. Payla\u015f\u0131ml\u0131 durum, \u00f6zellikle global<\/code> de\u011fi\u015fkenler veya s\u0131n\u0131f \u00f6zelliklerinde s\u0131kl\u0131kla kar\u015f\u0131m\u0131za \u00e7\u0131kar. A\u015fa\u011f\u0131daki basit kod \u00f6rne\u011fi, hen\u00fcz bir yar\u0131\u015f durumu olu\u015fturmadan \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n temel yap\u0131s\u0131n\u0131 g\u00f6stermektedir:<\/p>\n

\n# Basit bir Python multithreading \u00f6rne\u011fi\nimport threading\nimport time\n\ndef is_yap(ad):\n    \"\"\"Belirtilen isimle bir i\u015f par\u00e7ac\u0131\u011f\u0131 g\u00f6revi sim\u00fcle eder.\"\"\"\n    print(f\"[{ad}]: Ba\u015flad\u0131.\")\n    time.sleep(1) # \u0130\u015flem s\u00fcresini sim\u00fcle etmek i\u00e7in k\u0131sa bir bekleme\n    print(f\"[{ad}]: Bitti.\")\n\n# \u0130\u015f par\u00e7ac\u0131klar\u0131n\u0131 olu\u015fturma ve ba\u015flatma\nthreads = []\nfor i in range(3):\n    thread = threading.Thread(target=is_yap, args=(f\"\u0130\u015f Par\u00e7ac\u0131\u011f\u0131-{i+1}\",))\n    threads.append(thread)\n    thread.start() # \u0130\u015f par\u00e7ac\u0131\u011f\u0131n\u0131 ba\u015flat\n\n# T\u00fcm i\u015f par\u00e7ac\u0131klar\u0131n\u0131n tamamlanmas\u0131n\u0131 bekleme\nfor thread in threads:\n    thread.join() # \u0130\u015f par\u00e7ac\u0131\u011f\u0131 bitene kadar bekle\n\nprint(\"T\u00fcm i\u015f par\u00e7ac\u0131klar\u0131 g\u00f6revlerini ba\u015far\u0131yla tamamlad\u0131.\")\n  <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnekte, her i\u015f par\u00e7ac\u0131\u011f\u0131 birbirinden ba\u011f\u0131ms\u0131z \u00e7al\u0131\u015f\u0131r ve payla\u015f\u0131ml\u0131 bir kayna\u011fa eri\u015fim yoktur, dolay\u0131s\u0131yla bir yar\u0131\u015f durumu olu\u015fmaz. Ancak ger\u00e7ek d\u00fcnya uygulamalar\u0131nda, i\u015f par\u00e7ac\u0131klar\u0131n\u0131n etkile\u015fimi ka\u00e7\u0131n\u0131lmazd\u0131r ve i\u015fte tam da bu noktada senkronizasyon mekanizmalar\u0131na ihtiya\u00e7 duyar\u0131z. Yar\u0131\u015f durumlar\u0131, program\u0131n mant\u0131\u011f\u0131n\u0131 alt\u00fcst edebilir, veri b\u00fct\u00fcnl\u00fc\u011f\u00fcn\u00fc bozabilir ve hatta g\u00fcvenlik a\u00e7\u0131klar\u0131na yol a\u00e7abilir. Bu nedenle, \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalar geli\u015ftirirken bu t\u00fcr potansiyel sorunlara kar\u015f\u0131 dikkatli olmak ve uygun \u00f6nlemleri almak hayati \u00f6neme sahiptir.<\/p>\n
Yar\u0131\u015f Durumu Senaryolar\u0131: Ger\u00e7ek D\u00fcnyada Bizi Nas\u0131l Etkiler?<\/h2>\nYar\u0131\u015f durumlar\u0131, sadece teorik bir kavram olmaktan \u00e7ok, g\u00fcnl\u00fck kulland\u0131\u011f\u0131m\u0131z bir\u00e7ok uygulamada kendini g\u00f6sterebilen somut hatalara yol a\u00e7ar. Bu hatalar genellikle \u00f6ng\u00f6r\u00fclemez olduklar\u0131 ve belirli zamanlama ko\u015fullar\u0131na ba\u011fl\u0131 olduklar\u0131 i\u00e7in tespit edilmesi ve d\u00fczeltilmesi olduk\u00e7a zordur. \u015eimdi, yar\u0131\u015f durumlar\u0131n\u0131n ger\u00e7ek d\u00fcnya senaryolar\u0131nda nas\u0131l ortaya \u00e7\u0131kt\u0131\u011f\u0131na dair birka\u00e7 vaka analizini ve basit bir Python kodu \u00f6rne\u011fi \u00fczerinden bu kaosu nas\u0131l deneyimleyebilece\u011fimizi inceleyelim.<\/p>\n
Vaka Analizi 1: Online Bankac\u0131l\u0131k Sistemi<\/strong><\/p>\n
Bir online bankac\u0131l\u0131k sisteminde, kullan\u0131c\u0131lar\u0131n hesaplar\u0131ndan ayn\u0131 anda para \u00e7ekme veya yat\u0131rma i\u015flemleri yapt\u0131\u011f\u0131n\u0131 d\u00fc\u015f\u00fcn\u00fcn. Diyelim ki, Ali'nin hesab\u0131nda 1000 TL var. Ali, mobil uygulamas\u0131ndan 700 TL \u00e7ekmek isterken, ayn\u0131 anda e\u015fi de ba\u015fka bir cihazdan 600 TL \u00e7ekmeye \u00e7al\u0131\u015f\u0131yor. E\u011fer sistemde uygun senkronizasyon mekanizmalar\u0131 yoksa, \u015fu senaryo ger\u00e7ekle\u015febilir:<\/p>\n
\n\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A (Ali'nin i\u015flemi) bakiyeyi okur: 1000 TL.<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B (E\u015finin i\u015flemi) bakiyeyi okur: 1000 TL.<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A 700 TL'yi d\u00fc\u015fer, yeni bakiye 300 TL olur.<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B 600 TL'yi d\u00fc\u015fer, yeni bakiye 400 TL olur.<\/li>\n<\/ol>\nHer iki i\u015f par\u00e7ac\u0131\u011f\u0131 da kendi hesaplamalar\u0131n\u0131 eski bakiye de\u011feri \u00fczerinden yapt\u0131\u011f\u0131 i\u00e7in, toplamda 1300 TL \u00e7ekilmesine ra\u011fmen sistemde kalan bakiye 400 TL olarak g\u00f6r\u00fcn\u00fcr. Ger\u00e7ekte bakiye eksi 300 TL olmal\u0131yd\u0131. Bu durum, \"eksi bakiye\" gibi finansal tutars\u0131zl\u0131klara ve ciddi g\u00fcvenlik a\u00e7\u0131klar\u0131na yol a\u00e7abilir.<\/p>\n
Vaka Analizi 2: E-ticaret Envanter Y\u00f6netimi<\/strong><\/p>\n
Pop\u00fcler bir e-ticaret sitesinde, bir \u00fcr\u00fcn\u00fcn stok adedinin 5 oldu\u011funu ve iki farkl\u0131 m\u00fc\u015fterinin ayn\u0131 anda bu \u00fcr\u00fcnden sipari\u015f vermeye \u00e7al\u0131\u015ft\u0131\u011f\u0131n\u0131 hayal edin. Her iki m\u00fc\u015fterinin de 3'er adet sipari\u015f etti\u011fini varsayal\u0131m:<\/p>\n
\n\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A (M\u00fc\u015fteri 1) stok adedini okur: 5. \u0130htiyac\u0131 olan 3 adet sto\u011fu kontrol eder (5 > 3, uygun).<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B (M\u00fc\u015fteri 2) stok adedini okur: 5. \u0130htiyac\u0131 olan 3 adet sto\u011fu kontrol eder (5 > 3, uygun).<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A, 3 adet \u00fcr\u00fcn\u00fc stoktan d\u00fc\u015fer, yeni stok: 2.<\/li>\n
\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B, 3 adet \u00fcr\u00fcn\u00fc stoktan d\u00fc\u015fer, yeni stok: 2.<\/li>\n<\/ol>\nSonu\u00e7 olarak, 6 adet \u00fcr\u00fcn sat\u0131lm\u0131\u015f gibi g\u00f6r\u00fcnse de, sistemde sadece 2 adet \u00fcr\u00fcn kalm\u0131\u015ft\u0131r. Bu da, 4 adet \"over-sell\" (fazla sat\u0131\u015f) durumu yarat\u0131r ve m\u00fc\u015fteriye \u00fcr\u00fcn g\u00f6nderilemez. Bu durum, m\u00fc\u015fteri memnuniyetsizli\u011fine, iptal edilen sipari\u015flere ve \u015firket i\u00e7in finansal kay\u0131plara neden olur.<\/p>\n
Python'da Yar\u0131\u015f Durumu Olu\u015fturan K\u00f6t\u00fc Bir \u00d6rnek<\/strong><\/p>\n
Yukar\u0131daki bankac\u0131l\u0131k senaryosunu Python kodu ile sim\u00fcle edelim. Burada bakiye<\/code> de\u011fi\u015fkeni payla\u015f\u0131ml\u0131 bir kaynakt\u0131r ve birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 ayn\u0131 anda ona eri\u015fmeye \u00e7al\u0131\u015facakt\u0131r. time.sleep(0.1)<\/code> \u00e7a\u011fr\u0131s\u0131, i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda ba\u011flam de\u011fi\u015ftirme (context switching) olas\u0131l\u0131\u011f\u0131n\u0131 art\u0131rarak yar\u0131\u015f durumunun daha belirgin hale gelmesine yard\u0131mc\u0131 olur:<\/p>\n
\n# Yar\u0131\u015f Durumu Olu\u015fturan K\u00f6t\u00fc Bir \u00d6rnek: Banka Hesab\u0131 Sim\u00fclasyonu\nimport threading\nimport time\n\nbakiye = 1000 # Payla\u015f\u0131ml\u0131 kaynak: Banka hesab\u0131 bakiyesi\n\ndef para_cek(miktar):\n    \"\"\"Hesaptan belirli bir miktar para \u00e7ekme i\u015flemi.\"\"\"\n    global bakiye # Global bakiye de\u011fi\u015fkenine eri\u015fiyoruz\n    if bakiye >= miktar:\n        print(f\"[{threading.current_thread().name}]: Bakiye kontrol edildi: {bakiye}. \u00c7ekilecek: {miktar}\")\n        time.sleep(0.1) # \u0130\u015flem s\u00fcresini sim\u00fcle et, ba\u011flam de\u011fi\u015ftirme \u015fans\u0131n\u0131 art\u0131r\u0131r\n        bakiye -= miktar\n        print(f\"[{threading.current_thread().name}]: Para \u00e7ekildi. Yeni bakiye: {bakiye}\")\n    else:\n        print(f\"[{threading.current_thread().name}]: Yetersiz bakiye. Mevcut: {bakiye}, \u00c7ekilmek istenen: {miktar}\")\n\n# \u0130ki i\u015f par\u00e7ac\u0131\u011f\u0131 olu\u015fturma ve ba\u015flatma\n# thread1: 700 TL \u00e7ekmek istiyor\n# thread2: 600 TL \u00e7ekmek istiyor\nthread1 = threading.Thread(target=para_cek, args=(700,), name=\"M\u00fc\u015fteri A\")\nthread2 = threading.Thread(target=para_cek, args=(600,), name=\"M\u00fc\u015fteri B\")\n\nthread1.start()\nthread2.start()\n\nthread1.join()\nthread2.join()\n\nprint(f\"\\nBeklenen son bakiye: 1000 - 700 - 600 = -300 TL (E\u011fer do\u011fru \u00e7al\u0131\u015fsayd\u0131)\")\nprint(f\"Ger\u00e7ekle\u015fen Son Bakiye: {bakiye} TL\") # \u00c7o\u011fu zaman beklenenden farkl\u0131 bir sonu\u00e7 al\u0131n\u0131r\n  <\/pre>\n<\/code><\/p>\n
Bu kodu \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131n\u0131zda, b\u00fcy\u00fck olas\u0131l\u0131kla son bakiyenin -300 TL yerine 300 TL veya 400 TL gibi hatal\u0131 bir de\u011fer oldu\u011funu g\u00f6receksiniz. Bunun nedeni, her iki i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n da bakiye >= miktar<\/code> kontrol\u00fcn\u00fc ayn\u0131 anda veya \u00e7ok yak\u0131n zamanlarda yapmas\u0131 ve her ikisinin de bakiyeyi d\u00fc\u015f\u00fcrmek i\u00e7in yeterli miktar oldu\u011funu d\u00fc\u015f\u00fcnmesidir. Bu da payla\u015f\u0131ml\u0131 kaynak olan bakiye<\/code> \u00fczerinde bir tutars\u0131zl\u0131\u011fa yol a\u00e7ar. Bu t\u00fcr durumlar, \"kritik b\u00f6l\u00fcm\" (critical section) olarak adland\u0131r\u0131lan, payla\u015f\u0131ml\u0131 kaynaklara eri\u015fim sa\u011flayan kod bloklar\u0131n\u0131n iyi y\u00f6netilmesi gerekti\u011fini a\u00e7\u0131k\u00e7a ortaya koymaktad\u0131r. \u0130\u015fte bu noktada senkronizasyon ara\u00e7lar\u0131 devreye girer.<\/p>\n
Python'da Yar\u0131\u015f Durumlar\u0131n\u0131 Kilitleme (Lock) Mekanizmalar\u0131yla Nas\u0131l \u00d6nleriz?<\/h2>\nYar\u0131\u015f durumlar\u0131n\u0131n sebep oldu\u011fu veri tutars\u0131zl\u0131klar\u0131n\u0131 ve hatalar\u0131 \u00f6nlemek i\u00e7in, Python'\u0131n threading<\/code> mod\u00fcl\u00fcnde \u00e7e\u015fitli senkronizasyon mekanizmalar\u0131 bulunur. Bu mekanizmalar\u0131n ba\u015f\u0131nda \"Kilitler\" (Locks) gelir. Kilitler, bir anda sadece bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n belirli bir kod blo\u011funa veya kayna\u011fa eri\u015fimini garanti eden basit ama etkili ara\u00e7lard\u0131r. Bu kod blo\u011funa, yani payla\u015f\u0131ml\u0131 kayna\u011fa eri\u015fim sa\u011flayan b\u00f6l\u00fcme, \"kritik b\u00f6l\u00fcm\" ad\u0131n\u0131 veririz. Bir i\u015f par\u00e7ac\u0131\u011f\u0131 kritik b\u00f6l\u00fcme girmek istedi\u011finde kilidi \"elde etmeye\" (acquire) \u00e7al\u0131\u015f\u0131r. E\u011fer kilit serbestse, i\u015f par\u00e7ac\u0131\u011f\u0131 kilidi al\u0131r ve kritik b\u00f6l\u00fcme girer. Kilit art\u0131k me\u015fguld\u00fcr. Ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131 ayn\u0131 anda kritik b\u00f6l\u00fcme girmek isterse, kilit serbest kalana kadar beklemek zorunda kal\u0131r. \u0130\u015f par\u00e7ac\u0131\u011f\u0131 kritik b\u00f6l\u00fcmden \u00e7\u0131kt\u0131\u011f\u0131nda, kilidi \"serbest b\u0131rak\u0131r\" (release) ve b\u00f6ylece di\u011fer bekleyen i\u015f par\u00e7ac\u0131klar\u0131n\u0131n eri\u015fimine izin verir. Bu y\u00f6ntem, payla\u015f\u0131ml\u0131 kaynaklara eri\u015fimin s\u0131ras\u0131n\u0131 d\u00fczenleyerek veri b\u00fct\u00fcnl\u00fc\u011f\u00fcn\u00fc sa\u011flar.<\/p>\n
Python'da temel kilit mekanizmas\u0131 threading.Lock<\/code> s\u0131n\u0131f\u0131d\u0131r. Bu s\u0131n\u0131f\u0131n iki ana metodu vard\u0131r: acquire()<\/code> kilidi elde etmek i\u00e7in, release()<\/code> ise kilidi serbest b\u0131rakmak i\u00e7in kullan\u0131l\u0131r. Daha g\u00fcvenli ve Pythonik bir yakla\u015f\u0131m ise, kilit nesnesini bir ba\u011flam y\u00f6neticisi olarak (with<\/code> ifadesiyle) kullanmakt\u0131r. Bu \u015fekilde, acquire()<\/code> ve release()<\/code> metodlar\u0131n\u0131 manuel olarak \u00e7a\u011f\u0131rman\u0131za gerek kalmaz; with<\/code> blo\u011funa girildi\u011finde kilit otomatik olarak elde edilir ve bloktan \u00e7\u0131k\u0131ld\u0131\u011f\u0131nda (hata olsa bile) otomatik olarak serbest b\u0131rak\u0131l\u0131r. Bu, hata riskini azalt\u0131r ve kodu daha okunabilir hale getirir.<\/p>\n
\n# Lock Kullan\u0131m\u0131 ile Yar\u0131\u015f Durumunu Engellemek\nimport threading\nimport time\n\nbakiye = 1000\nkilit = threading.Lock() # Kilit nesnesi olu\u015fturuldu\n\ndef guvenli_para_cek(miktar):\n    \"\"\"Hesaptan g\u00fcvenli bir \u015fekilde para \u00e7ekme i\u015flemi.\"\"\"\n    global bakiye\n    # 'with kilit:' ifadesi, kilit.acquire() ve kilit.release() \u00e7a\u011fr\u0131lar\u0131n\u0131 otomatik y\u00f6netir.\n    # Bu, kritik b\u00f6l\u00fcm\u00fcn sadece bir i\u015f par\u00e7ac\u0131\u011f\u0131 taraf\u0131ndan eri\u015filmesini sa\u011flar.\n    with kilit: \n        if bakiye >= miktar:\n            print(f\"[{threading.current_thread().name}]: Bakiye kontrol edildi: {bakiye}. \u00c7ekilecek: {miktar}\")\n            time.sleep(0.1) # \u0130\u015flem s\u00fcresini sim\u00fcle et\n            bakiye -= miktar\n            print(f\"[{threading.current_thread().name}]: Para \u00e7ekildi. Yeni bakiye: {bakiye}\")\n        else:\n            print(f\"[{threading.current_thread().name}]: Yetersiz bakiye. Mevcut: {bakiye}, \u00c7ekilmek istenen: {miktar}\")\n\nthread1 = threading.Thread(target=guvenli_para_cek, args=(700,), name=\"G\u00fcvenli M\u00fc\u015fteri A\")\nthread2 = threading.Thread(target=guvenli_para_cek, args=(600,), name=\"G\u00fcvenli M\u00fc\u015fteri B\")\n\nthread1.start()\nthread2.start()\n\nthread1.join()\nthread2.join()\n\nprint(f\"\\nBeklenen son bakiye: 1000 - 700 - 600 = -300 TL\")\nprint(f\"Ger\u00e7ekle\u015fen Son Bakiye: {bakiye} TL\") # Bu sefer do\u011fru sonu\u00e7 al\u0131nmal\u0131\n  <\/pre>\n<\/code><\/p>\n
Bu kodu \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131n\u0131zda, sonu\u00e7 bakiye her zaman -300 TL olacakt\u0131r. \u00c7\u00fcnk\u00fc kilit<\/code> nesnesi, guvenli_para_cek<\/code> fonksiyonunun kritik b\u00f6l\u00fcm\u00fcn\u00fc ayn\u0131 anda yaln\u0131zca bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n y\u00fcr\u00fctmesine izin verir. \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A bakiyeyi kontrol edip \u00e7ekerken, \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B bekler. \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A bitip kilidi serbest b\u0131rakt\u0131\u011f\u0131nda, \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B kilidi al\u0131r ve kendi i\u015flemini yapar, ancak bu sefer g\u00fcncel bakiyeyi okudu\u011fu i\u00e7in i\u015flem ya do\u011fru bir \u015fekilde tamamlan\u0131r ya da yetersiz bakiye nedeniyle reddedilir. Bu y\u00f6ntem, verilerin tutarl\u0131 kalmas\u0131n\u0131 garanti eder.<\/p>\n
\n    Uzman \u0130pucu: Lock<\/code> kullan\u0131rken dikkatli olun! E\u011fer kilidi acquire<\/code> ettikten sonra release<\/code> etmeyi unutursan\u0131z, di\u011fer t\u00fcm i\u015f par\u00e7ac\u0131klar\u0131 s\u00fcresiz olarak bekleyebilir ve uygulaman\u0131z \"kilitlenmeye\" (deadlock) girebilir. with<\/code> ifadesi bu t\u00fcr hatalar\u0131 b\u00fcy\u00fck \u00f6l\u00e7\u00fcde \u00f6nler, bu y\u00fczden her zaman tercih edilmelidir. Kilitlerinizi m\u00fcmk\u00fcn olan en dar kapsamda (sadece payla\u015f\u0131ml\u0131 kayna\u011fa eri\u015fti\u011finiz kod blo\u011fu) tutmak, kilit \u00e7eki\u015fmesini (lock contention) azaltarak performans\u0131 art\u0131r\u0131r.
\n  <\/aside>\nYeniden Girilebilir Kilitler: threading.RLock<\/code><\/strong><\/p>\n
Bazen, bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n zaten tuttu\u011fu bir kilidi tekrar elde etmesi gerekebilir. threading.Lock<\/code> nesnesi bu duruma izin vermez; ayn\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131 ikinci kez acquire()<\/code> \u00e7a\u011fr\u0131s\u0131 yaparsa kilitlenir. \u0130\u015fte bu senaryolar i\u00e7in threading.RLock<\/code> (Re-entrant Lock - Yeniden Girilebilir Kilit) kullan\u0131\u015fl\u0131d\u0131r. RLock<\/code>, ayn\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kilidi birden fazla kez elde etmesine izin verir, ancak her acquire()<\/code> \u00e7a\u011fr\u0131s\u0131na kar\u015f\u0131l\u0131k gelen bir release()<\/code> \u00e7a\u011fr\u0131s\u0131 olmal\u0131d\u0131r. Yani, kilit serbest b\u0131rak\u0131lmadan \u00f6nce ka\u00e7 kez elde edildiyse, o kadar kez serbest b\u0131rak\u0131lmal\u0131d\u0131r. Bu, \u00f6zellikle i\u00e7 i\u00e7e fonksiyon \u00e7a\u011fr\u0131lar\u0131nda veya karma\u015f\u0131k veri yap\u0131lar\u0131nda ayn\u0131 kilit mekanizmas\u0131n\u0131n yeniden kullan\u0131lmas\u0131 gerekti\u011finde faydal\u0131d\u0131r.<\/p>\n
\u00d6zetle, threading.Lock<\/code> ve threading.RLock<\/code> gibi kilit mekanizmalar\u0131, \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 ortam\u0131nda payla\u015f\u0131ml\u0131 kaynaklara g\u00fcvenli ve tutarl\u0131 eri\u015fimi garanti etmenin temel yoludur. Do\u011fru kullan\u0131ld\u0131klar\u0131nda, yar\u0131\u015f durumlar\u0131n\u0131n neden oldu\u011fu kaosun \u00f6n\u00fcne ge\u00e7er ve uygulamalar\u0131n\u0131z\u0131n sa\u011flaml\u0131\u011f\u0131n\u0131 art\u0131r\u0131rlar. Ancak her senkronizasyon arac\u0131 gibi, bunlar\u0131n da performans maliyetleri ve potansiyel kilitlenme riskleri oldu\u011funu unutmamak \u00f6nemlidir.<\/p>\n
Daha Geli\u015fmi\u015f Senkronizasyon Ara\u00e7lar\u0131: Semaphore ve Condition De\u011fi\u015fkenleri Ne \u0130\u015fe Yarar?<\/h2>\nKilitler (Locks) tek bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kritik bir b\u00f6l\u00fcme eri\u015fimini sa\u011flarken, daha karma\u015f\u0131k senkronizasyon ihtiya\u00e7lar\u0131 i\u00e7in ba\u015fka ara\u00e7lara da ihtiya\u00e7 duyulur. Bu ara\u00e7lardan ikisi threading.Semaphore<\/code> ve threading.Condition<\/code> de\u011fi\u015fkenleridir. Her ikisi de farkl\u0131 senaryolarda kritik \u00f6neme sahip olup, \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 y\u00f6netimini daha esnek hale getirir.<\/p>\n
Semaphore: Kaynak Havuzlar\u0131n\u0131 Y\u00f6netmek<\/h3>\nthreading.Semaphore<\/code>, belirli bir say\u0131da i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n ayn\u0131 anda bir kayna\u011fa eri\u015fimine izin vermek i\u00e7in kullan\u0131lan bir senkronizasyon ilkelidir. Bir kilit (Lock) yaln\u0131zca tek bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n eri\u015fimine izin verirken, bir Semaphore belirtilen say\u0131da (say\u0131m de\u011feriyle ba\u015flat\u0131l\u0131r) i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n eri\u015fimine izin verir. Bu, \u00f6zellikle s\u0131n\u0131rl\u0131 kaynaklara (\u00f6rne\u011fin, veritaban\u0131 ba\u011flant\u0131 havuzlar\u0131, API \u00e7a\u011fr\u0131 limitleri veya i\u015f par\u00e7ac\u0131\u011f\u0131 havuzlar\u0131) sahip oldu\u011funuz senaryolarda \u00e7ok kullan\u0131\u015fl\u0131d\u0131r. Semaphore, ba\u015flang\u0131\u00e7ta pozitif bir tam say\u0131 de\u011feri (say\u0131mc\u0131) ile ba\u015flat\u0131l\u0131r. acquire()<\/code> \u00e7a\u011fr\u0131s\u0131 yap\u0131ld\u0131\u011f\u0131nda say\u0131mc\u0131 bir azalt\u0131l\u0131r, release()<\/code> \u00e7a\u011fr\u0131s\u0131 yap\u0131ld\u0131\u011f\u0131nda ise bir art\u0131r\u0131l\u0131r. Say\u0131mc\u0131 s\u0131f\u0131r oldu\u011funda, ba\u015fka hi\u00e7bir i\u015f par\u00e7ac\u0131\u011f\u0131 acquire()<\/code> \u00e7a\u011fr\u0131s\u0131 yapamaz ve say\u0131mc\u0131 tekrar pozitif bir de\u011fer alana kadar bekler.<\/p>\n
\n# Semaphore Kullan\u0131m\u0131na Bir \u00d6rnek: S\u0131n\u0131rl\u0131 Kaynak Havuzu Y\u00f6netimi\nimport threading\nimport time\nimport random\n\n# Ayn\u0131 anda en fazla 3 i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kayna\u011f\u0131 kullanmas\u0131na izin veren Semaphore\nkaynak_havuzu = threading.Semaphore(3) \n\ndef kaynak_kullan(thread_id):\n    \"\"\"Bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n s\u0131n\u0131rl\u0131 bir kayna\u011f\u0131 kullanmas\u0131n\u0131 sim\u00fcle eder.\"\"\"\n    print(f\"\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 {thread_id}: Kaynak bekliyor...\")\n    kaynak_havuzu.acquire() # Kayna\u011f\u0131 elde etmeye \u00e7al\u0131\u015f, say\u0131mc\u0131 0 ise bekle\n    try:\n        print(f\"\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 {thread_id}: Kayna\u011f\u0131 kullanmaya ba\u015flad\u0131.\")\n        time.sleep(random.uniform(0.5, 2.0)) # Kaynak kullan\u0131m s\u00fcresi sim\u00fclasyonu\n        print(f\"\u0130\u015f Par\u00e7ac\u0131\u011f\u0131 {thread_id}: Kayna\u011f\u0131 kullanmay\u0131 bitirdi.\")\n    finally:\n        kaynak_havuzu.release() # Kayna\u011f\u0131 serbest b\u0131rak, say\u0131mc\u0131y\u0131 art\u0131r\n\nthreads = []\nfor i in range(10): # 10 farkl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131 olu\u015fturuyoruz\n    thread = threading.Thread(target=kaynak_kullan, args=(i+1,))\n    threads.append(thread)\n    thread.start()\n\nfor thread in threads:\n    thread.join()\nprint(\"T\u00fcm kaynak kullan\u0131mlar\u0131 tamamland\u0131.\")\n  <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnekte, ayn\u0131 anda en fazla 3 i\u015f par\u00e7ac\u0131\u011f\u0131 \"kayna\u011f\u0131 kullan\u0131yor\" mesaj\u0131n\u0131 verecek ve di\u011ferleri bekleyecektir. Bu, bir veritaban\u0131 ba\u011flant\u0131 havuzunu veya bir web hizmetinin API \u00e7a\u011fr\u0131 limitini y\u00f6netmek i\u00e7in ideal bir y\u00f6ntemdir. Semaphore, kaynaklar\u0131n a\u015f\u0131r\u0131 kullan\u0131m\u0131n\u0131 \u00f6nleyerek sistemin stabil kalmas\u0131na yard\u0131mc\u0131 olur ve potansiyel performans sorunlar\u0131n\u0131 minimize eder.<\/p>\n
\n    Uzman \u0130pucu: Semaphore<\/code> kullan\u0131m\u0131, \u00f6zellikle \u00fcretici-t\u00fcketici desenlerinde, bir kuyru\u011fun maksimum boyutunu s\u0131n\u0131rlamak veya i\u015flenebilecek g\u00f6rev say\u0131s\u0131n\u0131 kontrol etmek i\u00e7in de \u00e7ok etkilidir. Bu sayede sistem kaynaklar\u0131 dengeli bir \u015fekilde kullan\u0131l\u0131r ve ani y\u00fcklenmelerin \u00f6n\u00fcne ge\u00e7ilir.
\n  <\/aside>\nCondition De\u011fi\u015fkenleri: \u0130\u015f Par\u00e7ac\u0131klar\u0131 Aras\u0131 \u0130leti\u015fim ve Bekleme Mekanizmalar\u0131<\/h3>\nthreading.Condition<\/code> de\u011fi\u015fkenleri, i\u015f par\u00e7ac\u0131klar\u0131n\u0131n belirli bir ko\u015ful do\u011fru olana kadar bekleyebilmesini ve bu ko\u015ful kar\u015f\u0131land\u0131\u011f\u0131nda ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131 taraf\u0131ndan uyand\u0131r\u0131lmas\u0131n\u0131 sa\u011flayan daha karma\u015f\u0131k bir senkronizasyon arac\u0131d\u0131r. Basit kilitlerden farkl\u0131 olarak, Condition de\u011fi\u015fkenleri i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda daha sofistike bir koordinasyon sa\u011flar. Bir Condition de\u011fi\u015fkeni her zaman bir Lock<\/code> veya RLock<\/code> ile ili\u015fkilidir. Temel \u00e7al\u0131\u015fma prensibi \u015f\u00f6yledir:<\/p>\n
\nBir i\u015f par\u00e7ac\u0131\u011f\u0131 bir ko\u015fulun ger\u00e7ekle\u015fmesini bekliyorsa, Condition de\u011fi\u015fkenini acquire()<\/code> eder, ko\u015fulu kontrol eder.<\/li>\n
Ko\u015ful hen\u00fcz sa\u011flanmam\u0131\u015fsa, wait()<\/code> metodunu \u00e7a\u011f\u0131r\u0131r. wait()<\/code> metodu, ili\u015fkilendirilmi\u015f kilidi otomatik olarak serbest b\u0131rak\u0131r ve i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131 uykuya dald\u0131r\u0131r.<\/li>\n
Ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131, ko\u015fulu de\u011fi\u015ftirecek bir i\u015flem yapt\u0131\u011f\u0131nda (yine Condition de\u011fi\u015fkenini acquire()<\/code> ederek), bu de\u011fi\u015fiklikten sonra notify()<\/code> veya notify_all()<\/code> metodunu \u00e7a\u011f\u0131rarak bekleyen i\u015f par\u00e7ac\u0131klar\u0131n\u0131 uyand\u0131r\u0131r.<\/li>\n
Uyand\u0131r\u0131lan i\u015f par\u00e7ac\u0131\u011f\u0131, wait()<\/code> metodundan geri d\u00f6nerken, daha \u00f6nce serbest b\u0131rakt\u0131\u011f\u0131 kilidi otomatik olarak tekrar elde eder ve ko\u015fulu tekrar kontrol eder.<\/li>\n<\/ol>\nBu mekanizma, \u00f6zellikle \"\u00fcretici-t\u00fcketici\" senaryolar\u0131nda, bir \u00fcreticinin veri \u00fcretti\u011finde bir t\u00fcketiciyi uyand\u0131rmas\u0131 veya bir t\u00fcketicinin veri olmad\u0131\u011f\u0131nda beklemeye ge\u00e7mesi gibi durumlar i\u00e7in idealdir. \u00d6rne\u011fin, bir veri kuyru\u011funda veri varken t\u00fcketici i\u015f par\u00e7ac\u0131klar\u0131n\u0131n \u00e7al\u0131\u015fmas\u0131n\u0131, veri yokken ise beklemeye ge\u00e7mesini sa\u011flamak i\u00e7in Condition de\u011fi\u015fkenlerini kullanabilirsiniz.<\/p>\n
\n# Condition De\u011fi\u015fkeni Kullan\u0131m\u0131: \u00dcretici-T\u00fcketici Senaryosu\nimport threading\nimport time\nimport random\n\nkuyruk = [] # Payla\u015f\u0131ml\u0131 veri kuyru\u011fu\nMAX_KUYRUK_BOYUTU = 5\n\n# Bir kilit ve onunla ili\u015fkilendirilmi\u015f bir Condition de\u011fi\u015fkeni olu\u015ftur\nkilit = threading.Lock()\ncondition = threading.Condition(kilit)\n\nclass Uretici(threading.Thread):\n    def run(self):\n        urun_sayisi = 10\n        for i in range(urun_sayisi):\n            with condition:\n                # Kuyruk doluysa bekle\n                while len(kuyruk) == MAX_KUYRUK_BOYUTU:\n                    print(f\"[\u00dcretici]: Kuyruk dolu, bekliyorum...\")\n                    condition.wait() # Kilidi serbest b\u0131rak ve uyanmay\u0131 bekle\n                \n                urun = f\"\u00dcr\u00fcn-{i+1}\"\n                kuyruk.append(urun)\n                print(f\"[\u00dcretici]: {urun} ekledi. Kuyruk: {kuyruk}\")\n                condition.notify_all() # Bekleyen t\u00fcketicileri uyand\u0131r\n                time.sleep(random.uniform(0.1, 0.5))\n        with condition:\n            kuyruk.append(None) # T\u00fcketicilere i\u015fin bitti\u011fini bildir\n            condition.notify_all()\n\n\nclass Tuketici(threading.Thread):\n    def run(self):\n        while True:\n            with condition:\n                # Kuyruk bo\u015fsa bekle\n                while not kuyruk:\n                    print(f\"[T\u00fcketici-{self.name}]: Kuyruk bo\u015f, bekliyorum...\")\n                    condition.wait() # Kilidi serbest b\u0131rak ve uyanmay\u0131 bekle\n\n                urun = kuyruk.pop(0) # \u0130lk eleman\u0131 al\n                if urun is None:\n                    kuyruk.append(None) # Di\u011fer t\u00fcketicilere de i\u015fin bitti\u011fini bildir\n                    condition.notify_all()\n                    break # T\u00fcketici i\u015fini bitirdi\n                \n                print(f\"[T\u00fcketici-{self.name}]: {urun} i\u015fliyorum. Kuyruk: {kuyruk}\")\n                condition.notify_all() # \u00dcreticiyi uyand\u0131r (yer a\u00e7\u0131ld\u0131)\n                time.sleep(random.uniform(0.5, 1.5))\n        print(f\"[T\u00fcketici-{self.name}]: \u0130\u015fini bitirdi.\")\n\n\nuretici = Uretici()\ntuketici1 = Tuketici(name=\"1\")\ntuketici2 = Tuketici(name=\"2\")\n\nuretici.start()\ntuketici1.start()\ntuketici2.start()\n\nuretici.join()\ntuketici1.join()\ntuketici2.join()\n\nprint(\"T\u00fcm \u00fcretim ve t\u00fcketim i\u015flemleri tamamland\u0131.\")\n  <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnekte, \u00fcretici kuyruk dolu oldu\u011funda beklerken, t\u00fcketiciler kuyruk bo\u015f oldu\u011funda beklerler. condition.notify_all()<\/code> \u00e7a\u011fr\u0131lar\u0131, ilgili i\u015f par\u00e7ac\u0131klar\u0131n\u0131 uyand\u0131rarak veri ak\u0131\u015f\u0131n\u0131 d\u00fczenler. Semaphore ve Condition de\u011fi\u015fkenleri, \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 uygulamalar\u0131nda daha karma\u015f\u0131k senkronizasyon ve koordinasyon ihtiya\u00e7lar\u0131n\u0131 kar\u015f\u0131lamak i\u00e7in vazge\u00e7ilmez ara\u00e7lard\u0131r. Bunlar, i\u015f par\u00e7ac\u0131klar\u0131n\u0131n kaynaklara eri\u015fimini ve birbirleriyle ileti\u015fimini daha d\u00fczenli ve \u00f6ng\u00f6r\u00fclebilir hale getirerek yar\u0131\u015f durumlar\u0131n\u0131n \u00f6n\u00fcne ge\u00e7er ve uygulaman\u0131n genel g\u00fcvenilirli\u011fini art\u0131r\u0131r.<\/p>\n
Python GIL (Global Interpreter Lock): Bir Engel mi, Bir Kalkan m\u0131?<\/h2>\nPython'da \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 (multithreading) konusunu incelerken, Global Interpreter Lock (GIL) kavram\u0131na de\u011finmemek olmaz. GIL, CPython (Python'\u0131n standart ve en yayg\u0131n kullan\u0131lan uygulamas\u0131) taraf\u0131ndan uygulanan benzersiz bir mekanizmad\u0131r ve \u00e7o\u011fu zaman hem bir engel hem de bir kalkan olarak alg\u0131lan\u0131r. Peki tam olarak nedir bu GIL ve yar\u0131\u015f durumlar\u0131yla ili\u015fkisi nas\u0131ld\u0131r?<\/p>\n
GIL Nedir?<\/strong><\/p>\n
Basit\u00e7e s\u00f6ylemek gerekirse, GIL, herhangi bir anda yaln\u0131zca bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n Python bayt kodunu y\u00fcr\u00fctmesine izin veren bir mutex'tir (kilittir). Yani, ayn\u0131 Python yorumlay\u0131c\u0131s\u0131 i\u00e7inde birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 olu\u015fturmu\u015f olsan\u0131z bile, yaln\u0131zca biri CPU \u00fczerinde ger\u00e7ekten \u00e7al\u0131\u015fabilir. Di\u011fer i\u015f par\u00e7ac\u0131klar\u0131, aktif i\u015f par\u00e7ac\u0131\u011f\u0131 GIL'i serbest b\u0131rakana kadar beklemek zorundad\u0131r. Bu k\u0131s\u0131tlama, CPython'\u0131n dahili veri yap\u0131lar\u0131n\u0131n (\u00f6rne\u011fin, referans say\u0131c\u0131lar gibi) i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli olmas\u0131n\u0131 sa\u011flamak i\u00e7in tasarlanm\u0131\u015ft\u0131r. Bu sayede, geli\u015ftiricilerin Python'da dahili olarak bir\u00e7ok yar\u0131\u015f durumuyla u\u011fra\u015fmas\u0131na gerek kalmaz.<\/p>\n
GIL Bir Kalkan m\u0131?<\/strong><\/p>\n
Evet, bir anlamda kalkan g\u00f6revi g\u00f6r\u00fcr. GIL sayesinde, Python'daki bir\u00e7ok C uzant\u0131s\u0131 (\u00f6rne\u011fin NumPy gibi k\u00fct\u00fcphaneler) veya temel veri yap\u0131lar\u0131, i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli olmak i\u00e7in karma\u015f\u0131k kilit mekanizmalar\u0131na ihtiya\u00e7 duymazlar. Bu, CPython yorumlay\u0131c\u0131s\u0131n\u0131n daha basit ve daha h\u0131zl\u0131 olmas\u0131n\u0131 sa\u011flar. Ayr\u0131ca, Python geli\u015ftiricileri olarak bizim, Python'\u0131n i\u00e7 i\u015fleyi\u015fi hakk\u0131nda endi\u015felenmek yerine, kendi uygulamalar\u0131m\u0131zdaki payla\u015f\u0131ml\u0131 verilere odaklanmam\u0131za olanak tan\u0131r. Yani, GIL, CPython'\u0131n dahili veri tutarl\u0131l\u0131\u011f\u0131n\u0131 sa\u011flamak i\u00e7in tasarlanm\u0131\u015f bir g\u00fcvenlik a\u011f\u0131d\u0131r.<\/p>\n
Peki GIL Bir Engel mi?<\/strong><\/p>\n
Kesinlikle evet. GIL'in en b\u00fcy\u00fck dezavantaj\u0131, Python'\u0131n ger\u00e7ek paralellikten (ayn\u0131 anda birden fazla CPU \u00e7ekirde\u011finde \u00e7al\u0131\u015fmaktan) yararlanmas\u0131n\u0131 engellemesidir. CPU-a\u011f\u0131rl\u0131kl\u0131 (CPU-bound) g\u00f6revler (yani, \u00e7o\u011fu zaman i\u015flemciyi kullanan ve I\/O beklemeyen g\u00f6revler) i\u00e7in birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 kullanmak, genellikle tek bir i\u015f par\u00e7ac\u0131\u011f\u0131 kullanmaktan daha h\u0131zl\u0131 sonu\u00e7 vermez, hatta GIL'in getirdi\u011fi ba\u011flam de\u011fi\u015ftirme (context switching) y\u00fck\u00fc nedeniyle daha yava\u015f bile olabilir. Bu, \u00e7ok \u00e7ekirdekli sistemlerin t\u00fcm g\u00fcc\u00fcn\u00fc kullanma yetene\u011fimizi s\u0131n\u0131rlar.<\/p>\n
GIL'in Etkisi: CPU-a\u011f\u0131rl\u0131kl\u0131 vs. I\/O-a\u011f\u0131rl\u0131kl\u0131 G\u00f6revler<\/strong><\/p>\n
\n\n      CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler (CPU-bound):<\/strong> Bu t\u00fcr g\u00f6revler, matematiksel hesaplamalar, g\u00f6r\u00fcnt\u00fc i\u015fleme veya yo\u011fun veri analizi gibi s\u00fcrekli i\u015flemciye ihtiya\u00e7 duyan i\u015flemlerdir. GIL, bu t\u00fcr g\u00f6revlerde birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n ayn\u0131 anda \u00e7al\u0131\u015fmas\u0131n\u0131 engelledi\u011fi i\u00e7in, bu i\u015f par\u00e7ac\u0131klar\u0131 asl\u0131nda s\u0131rayla y\u00fcr\u00fct\u00fcl\u00fcr. Sonu\u00e7 olarak, multithreading CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revlerde performans\u0131 art\u0131rmak yerine d\u00fc\u015f\u00fcrebilir.\n    <\/li>\n
\n      I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler (I\/O-bound):<\/strong> Bu t\u00fcr g\u00f6revler, dosya okuma\/yazma, a\u011f istekleri (HTTP \u00e7a\u011fr\u0131lar\u0131, veritaban\u0131 sorgular\u0131) veya kullan\u0131c\u0131 giri\u015fi gibi, \u00e7o\u011fu zaman bir i\u015flemden veri bekleyen i\u015flemlerdir. Bir i\u015f par\u00e7ac\u0131\u011f\u0131 bir I\/O i\u015flemi beklerken, GIL'i serbest b\u0131rak\u0131r ve b\u00f6ylece di\u011fer i\u015f par\u00e7ac\u0131klar\u0131n\u0131n CPU'ya eri\u015fmesine izin verir. Bu durumda, multithreading performans\u0131 \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131rabilir \u00e7\u00fcnk\u00fc bir i\u015f par\u00e7ac\u0131\u011f\u0131 beklerken di\u011ferleri \u00e7al\u0131\u015fmaya devam edebilir.\n    <\/li>\n<\/ul>\nA\u015fa\u011f\u0131daki kod \u00f6rne\u011fi, GIL'in CPU-a\u011f\u0131rl\u0131kl\u0131 ve I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler \u00fczerindeki etkisini g\u00f6stermektedir:<\/p>\n
\n# GIL'in CPU-a\u011f\u0131rl\u0131kl\u0131 ve I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revlere etkisi\nimport threading\nimport time\n\ndef cpu_agirlikli_islem():\n    \"\"\"Yo\u011fun bir CPU i\u015flemi sim\u00fclasyonu.\"\"\"\n    count = 0\n    # B\u00fcy\u00fck bir d\u00f6ng\u00fc ile i\u015flemciyi me\u015fgul et\n    for _ in range(5 * 10**7): # 50 milyon i\u015flem\n        count += 1\n    # print(f\"[{threading.current_thread().name}]: CPU i\u015flemi bitti.\")\n\ndef io_agirlikli_islem():\n    \"\"\"Bir I\/O beklemesi sim\u00fclasyonu.\"\"\"\n    # print(f\"[{threading.current_thread().name}]: I\/O i\u015flemi ba\u015flad\u0131.\")\n    time.sleep(1) # 1 saniyelik I\/O beklemesi\n    # print(f\"[{threading.current_thread().name}]: I\/O i\u015flemi bitti.\")\n\nprint(\"--- CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler ile test ---\")\nstart_time = time.time()\nt1 = threading.Thread(target=cpu_agirlikli_islem, name=\"CPU Thread 1\")\nt2 = threading.Thread(target=cpu_agirlikli_islem, name=\"CPU Thread 2\")\nt1.start()\nt2.start()\nt1.join()\nt2.join()\nprint(f\"\u0130ki CPU-a\u011f\u0131rl\u0131kl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131 s\u00fcresi: {time.time() - start_time:.4f} saniye\")\n# Genellikle tek i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n iki kat\u0131 s\u00fcreden biraz daha uzun s\u00fcrer (GIL nedeniyle)\n\nprint(\"\\n--- I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler ile test ---\")\nstart_time = time.time()\nt3 = threading.Thread(target=io_agirlikli_islem, name=\"IO Thread 1\")\nt4 = threading.Thread(target=io_agirlikli_islem, name=\"IO Thread 2\")\nt3.start()\nt4.start()\nt3.join()\nt4.join()\nprint(f\"\u0130ki I\/O-a\u011f\u0131rl\u0131kl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131 s\u00fcresi: {time.time() - start_time:.4f} saniye\")\n# Genellikle tek i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n s\u00fcresine yak\u0131n s\u00fcrer (GIL serbest b\u0131rak\u0131ld\u0131\u011f\u0131 i\u00e7in)\n  <\/pre>\n<\/code><\/p>\n
Yukar\u0131daki kodu \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131n\u0131zda, iki CPU-a\u011f\u0131rl\u0131kl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n toplam s\u00fcresinin, tek bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n s\u00fcresinin yakla\u015f\u0131k iki kat\u0131 oldu\u011funu, ancak iki I\/O-a\u011f\u0131rl\u0131kl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n toplam s\u00fcresinin, tek bir I\/O-a\u011f\u0131rl\u0131kl\u0131 i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n s\u00fcresine yak\u0131n (yakla\u015f\u0131k 1 saniye) oldu\u011funu g\u00f6receksiniz. Bu, GIL'in I\/O i\u015flemleri s\u0131ras\u0131nda serbest b\u0131rak\u0131lmas\u0131n\u0131n ve bu sayede e\u015f zamanl\u0131l\u0131\u011f\u0131n sa\u011flanmas\u0131n\u0131n somut bir g\u00f6stergesidir.<\/p>\n
GIL, Yar\u0131\u015f Durumlar\u0131n\u0131 Engeller mi?<\/strong><\/p>\n
GIL, CPython'\u0131n dahili veri yap\u0131lar\u0131 \u00fczerindeki yar\u0131\u015f durumlar\u0131n\u0131 engellerken, sizin uygulaman\u0131zdaki payla\u015f\u0131ml\u0131 Python nesneleri \u00fczerindeki yar\u0131\u015f durumlar\u0131n\u0131 (makalenin ba\u015f\u0131nda bahsetti\u011fimiz bakiye<\/code> \u00f6rne\u011fi gibi) otomatik olarak engellemez. \u00d6rne\u011fin, x += 1<\/code> i\u015flemi Python'da atomik de\u011fildir; bu asl\u0131nda \u00fc\u00e7 ayr\u0131 ad\u0131mdan olu\u015fur: x<\/code>'in de\u011ferini oku, de\u011feri 1 art\u0131r, yeni de\u011feri x<\/code>'e yaz. GIL, bu \u00fc\u00e7 ad\u0131m\u0131n birini yaparken ba\u015fka bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n m\u00fcdahale etmesini engelleyebilir, ancak t\u00fcm bu zincirin bir b\u00fct\u00fcn olarak atomik olmas\u0131n\u0131 sa\u011flamaz. Dolay\u0131s\u0131yla, payla\u015f\u0131ml\u0131 Python de\u011fi\u015fkenleri \u00fczerinde hala kilitler gibi harici senkronizasyon mekanizmalar\u0131na ihtiyac\u0131n\u0131z vard\u0131r.<\/p>\n
Sonu\u00e7 olarak, GIL, Python'\u0131n C uzant\u0131lar\u0131n\u0131n ve dahili yap\u0131lar\u0131n\u0131n g\u00fcvenli\u011fini sa\u011flamak i\u00e7in gerekli bir kalkan olsa da, ayn\u0131 zamanda CPU-a\u011f\u0131rl\u0131kl\u0131 paralel i\u015flem yeteneklerimizi k\u0131s\u0131tlayan bir engeldir. \u00c7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 kullan\u0131rken, GIL'in varl\u0131\u011f\u0131n\u0131 g\u00f6z \u00f6n\u00fcnde bulundurarak, uygulamalar\u0131n\u0131z\u0131n I\/O-a\u011f\u0131rl\u0131kl\u0131 m\u0131 yoksa CPU-a\u011f\u0131rl\u0131kl\u0131 m\u0131 oldu\u011funu belirlemeli ve buna g\u00f6re stratejiler geli\u015ftirmelisiniz. CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler i\u00e7in genellikle multiprocessing<\/code> mod\u00fcl\u00fc (her i\u015flemcinin kendi Python yorumlay\u0131c\u0131s\u0131 ve kendi GIL'i oldu\u011fu i\u00e7in ger\u00e7ek paralellik sa\u011flar) daha uygun bir \u00e7\u00f6z\u00fcm olacakt\u0131r.<\/p>\n
Veri Yap\u0131lar\u0131n\u0131 G\u00fcvenli Hale Getirmek: Queue<\/code> ve Atomik \u0130\u015flemler<\/h2>\n\u00c7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalarda en b\u00fcy\u00fck zorluklardan biri, i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda veri al\u0131\u015fveri\u015fini g\u00fcvenli ve verimli bir \u015fekilde yapmakt\u0131r. Payla\u015f\u0131ml\u0131 listeler, s\u00f6zl\u00fckler veya di\u011fer veri yap\u0131lar\u0131 \u00fczerinde do\u011frudan i\u015flem yapmak, yar\u0131\u015f durumlar\u0131na davetiye \u00e7\u0131kar\u0131r. Ancak Python, bu t\u00fcr senaryolar i\u00e7in \u00f6zel olarak tasarlanm\u0131\u015f \"thread-safe\" (i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli) veri yap\u0131lar\u0131 sunar. Bu yap\u0131lar\u0131n en bilineni ve yayg\u0131n olarak kullan\u0131lan\u0131 queue<\/code> mod\u00fcl\u00fcd\u00fcr.<\/p>\n
queue.Queue<\/code>: \u00dcretici-T\u00fcketici Deseni i\u00e7in M\u00fckemmel Bir \u00c7\u00f6z\u00fcm<\/h3>\nPython'\u0131n standart k\u00fct\u00fcphanesindeki queue<\/code> mod\u00fcl\u00fc, i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda g\u00fcvenli veri al\u0131\u015fveri\u015fi i\u00e7in bir dizi FIFO (First-In, First-Out - \u0130lk Giren \u0130lk \u00c7\u0131kar) kuyruk s\u0131n\u0131f\u0131 sa\u011flar. queue.Queue<\/code> s\u0131n\u0131f\u0131, temel bir Lock ve Condition de\u011fi\u015fkenlerini kendi i\u00e7inde bar\u0131nd\u0131rarak, put()<\/code> ve get()<\/code> gibi metodlar\u0131n\u0131 otomatik olarak senkronize eder. Bu sayede, siz manuel olarak kilit y\u00f6netimi yapmak zorunda kalmadan, verilerin kuyru\u011fa eklenmesi (\u00fcretici) ve kuyruktan al\u0131nmas\u0131 (t\u00fcketici) i\u015flemlerini g\u00fcvenle ger\u00e7ekle\u015ftirebilirsiniz. Bu, \"\u00fcretici-t\u00fcketici\" (producer-consumer) deseni i\u00e7in ideal bir \u00e7\u00f6z\u00fcmd\u00fcr, \u00e7\u00fcnk\u00fc bir veya daha fazla i\u015f par\u00e7ac\u0131\u011f\u0131 veri \u00fcretirken, bir veya daha fazla i\u015f par\u00e7ac\u0131\u011f\u0131 bu veriyi g\u00fcvenli bir \u015fekilde t\u00fcketebilir.<\/p>\n
\n# queue.Queue ile Thread-Safe \u0130leti\u015fim\nimport threading\nimport queue # queue mod\u00fcl\u00fcn\u00fc i\u00e7e aktar\u0131n\nimport time\nimport random\n\n# \u0130\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli bir kuyruk olu\u015fturun\nveri_kuyrugu = queue.Queue()\n\ndef uretici(kuyruk, urun_sayisi):\n    \"\"\"Belirtilen say\u0131da \u00fcr\u00fcn \u00fcreten i\u015f par\u00e7ac\u0131\u011f\u0131.\"\"\"\n    for i in range(urun_sayisi):\n        urun = f\"\u00dcr\u00fcn-{i+1}\"\n        time.sleep(random.uniform(0.1, 0.5)) # \u00dcretim s\u00fcresini sim\u00fcle et\n        kuyruk.put(urun) # Kuyru\u011fa g\u00fcvenli bir \u015fekilde \u00fcr\u00fcn ekle\n        print(f\"[{threading.current_thread().name}]: '{urun}' ekledi. Kuyruk boyutu: {kuyruk.qsize()}\")\n    # T\u00fcketicilere i\u015flerin bitti\u011fini bildirmek i\u00e7in \u00f6zel bir i\u015faret g\u00f6nderin\n    kuyruk.put(None) \n    print(f\"[{threading.current_thread().name}]: T\u00fcm \u00fcr\u00fcnleri \u00fcretti ve biti\u015f sinyali g\u00f6nderdi.\")\n\ndef tuketici(kuyruk, tuketici_id):\n    \"\"\"Kuyruktan \u00fcr\u00fcnleri t\u00fcketen i\u015f par\u00e7ac\u0131\u011f\u0131.\"\"\"\n    while True:\n        urun = kuyruk.get() # Kuyruktan g\u00fcvenli bir \u015fekilde \u00fcr\u00fcn al (bo\u015fsa bekler)\n        if urun is None:\n            # \u0130\u015f biti\u015f sinyalini ald\u0131k, di\u011fer t\u00fcketicilere de iletmek i\u00e7in geri koy\n            kuyruk.put(None)\n            break\n        print(f\"[{threading.current_thread().name} - T\u00fcketici {tuketici_id}]: '{urun}' i\u015fliyorum...\")\n        time.sleep(random.uniform(0.2, 0.8)) # \u0130\u015flem s\u00fcresini sim\u00fcle et\n        kuyruk.task_done() # \u0130\u015flem bitti\u011fini kuyru\u011fa bildir\n    print(f\"[{threading.current_thread().name} - T\u00fcketici {tuketici_id}]: \u0130\u015fini bitirdi.\")\n\n# \u00dcretici ve t\u00fcketici i\u015f par\u00e7ac\u0131klar\u0131n\u0131 olu\u015ftur\nuretici_thread = threading.Thread(target=uretici, args=(veri_kuyrugu, 10), name=\"\u00dcretici\")\ntuketici_threads = []\nfor i in range(2):\n    t = threading.Thread(target=tuketici, args=(veri_kuyrugu, i+1), name=f\"T\u00fcketici {i+1}\")\n    tuketici_threads.append(t)\n\n# \u0130\u015f par\u00e7ac\u0131klar\u0131n\u0131 ba\u015flat\nuretici_thread.start()\nfor t in tuketici_threads:\n    t.start()\n\n# \u00dcreticinin bitmesini bekle\nuretici_thread.join()\n# T\u00fcm t\u00fcketicilerin bitmesini bekle\nfor t in tuketici_threads:\n    t.join()\n\n# Kuyruktaki t\u00fcm g\u00f6revlerin tamamland\u0131\u011f\u0131n\u0131 onaylamak i\u00e7in\nveri_kuyrugu.join() \nprint(\"T\u00fcm \u00fcretim ve t\u00fcketim i\u015flemleri tamamland\u0131.\")\n  <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnekte, queue.Queue<\/code>'nun put()<\/code> ve get()<\/code> metodlar\u0131 i\u00e7sel olarak kilitler kullanarak yar\u0131\u015f durumlar\u0131n\u0131 engeller. E\u011fer kuyruk bo\u015fsa get()<\/code> \u00e7a\u011f\u0131ran i\u015f par\u00e7ac\u0131\u011f\u0131 otomatik olarak beklemeye al\u0131n\u0131r; e\u011fer kuyruk doluysa (ki Queue<\/code> varsay\u0131lan olarak s\u0131n\u0131rs\u0131zd\u0131r, ama queue.Queue(maxsize=...)<\/code> ile s\u0131n\u0131rlanabilir) put()<\/code> \u00e7a\u011f\u0131ran i\u015f par\u00e7ac\u0131\u011f\u0131 bekleyebilir. task_done()<\/code> ve join()<\/code> metodlar\u0131, bir i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kuyruktaki t\u00fcm \u00f6\u011felerin i\u015flenmesini beklemesi gerekti\u011fi durumlarda \u00e7ok kullan\u0131\u015fl\u0131d\u0131r. Bu, uygulaman\u0131z\u0131n kapanmadan \u00f6nce t\u00fcm i\u015flerin tamamland\u0131\u011f\u0131ndan emin olman\u0131z\u0131 sa\u011flar.<\/p>\n
Di\u011fer Thread-Safe Veri Yap\u0131lar\u0131 ve Atomik \u0130\u015flemler<\/h3>\nPython'\u0131n collections<\/code> mod\u00fcl\u00fcnde bulunan deque<\/code> (double-ended queue) da bir ba\u015fka kullan\u0131\u015fl\u0131 veri yap\u0131s\u0131d\u0131r. deque<\/code>'nun bir\u00e7ok i\u015flemi (append, popleft, popright gibi) atomik olarak kabul edilir, yani tek bir ad\u0131mda tamamlan\u0131r ve kesintiye u\u011framaz. Ancak, deque<\/code> \u00fczerinde birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 ayn\u0131 anda karma\u015f\u0131k i\u015flemler (\u00f6rne\u011fin, bir eleman\u0131 okuyup sonra silme) yaparken yine de kilit mekanizmalar\u0131na ihtiya\u00e7 duyulabilir.<\/p>\n
Atomik i\u015flemler, bir i\u015flemin b\u00f6l\u00fcnemez oldu\u011fu ve ya tamamen yap\u0131ld\u0131\u011f\u0131 ya da hi\u00e7 yap\u0131lmad\u0131\u011f\u0131 anlam\u0131na gelir. Python'da, bir\u00e7ok temel i\u015flem (\u00f6rne\u011fin, bir de\u011fi\u015fken atamas\u0131) genellikle atomik kabul edilir, ancak karma\u015f\u0131k i\u015flemler (\u00f6rne\u011fin, x += 1<\/code> gibi okuma-de\u011fi\u015ftirme-yazma d\u00f6ng\u00fcleri) atomik de\u011fildir ve bu nedenle yar\u0131\u015f durumlar\u0131na a\u00e7\u0131kt\u0131r. Python'da \u00f6zel atomik i\u015flem s\u0131n\u0131flar\u0131 olmasa da, threading.Lock<\/code> ve queue.Queue<\/code> gibi yap\u0131lar, bu karma\u015f\u0131k i\u015flemleri \"atomik\" hale getirerek veri g\u00fcvenli\u011fini sa\u011flama y\u00f6ntemleridir.<\/p>\n
\u00d6zetle, queue.Queue<\/code> gibi i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli veri yap\u0131lar\u0131, \u00f6zellikle \u00fcretici-t\u00fcketici desenleri gibi ortak kullan\u0131m durumlar\u0131nda yar\u0131\u015f durumlar\u0131ndan ka\u00e7\u0131nmak i\u00e7in paha bi\u00e7ilmez ara\u00e7lard\u0131r. Kendi kilit mekanizmalar\u0131n\u0131z\u0131 y\u00f6netme y\u00fck\u00fcn\u00fc \u00fczerinizden alarak, uygulaman\u0131z\u0131n kodunu daha temiz, daha g\u00fcvenli ve hata ay\u0131klamas\u0131 daha kolay hale getirirler. \u0130\u015f par\u00e7ac\u0131kl\u0131 uygulamalar geli\u015ftirirken, m\u00fcmk\u00fcn oldu\u011funca bu t\u00fcr haz\u0131r ve g\u00fcvenli yap\u0131lar\u0131 tercih etmek, ba\u015f a\u011fr\u0131lar\u0131n\u0131z\u0131n \u00e7o\u011funu giderecektir.<\/p>\n
Performans ve \u00d6l\u00e7eklenebilirlik: Yar\u0131\u015f Durumlar\u0131n\u0131 \u00d6nlerken Nelere Dikkat Etmeli?<\/h2>\nYar\u0131\u015f durumlar\u0131n\u0131 \u00f6nlemek i\u00e7in senkronizasyon mekanizmalar\u0131 kullanmak hayati \u00f6nem ta\u015f\u0131r, ancak bu mekanizmalar\u0131n beraberinde getirdi\u011fi baz\u0131 performans maliyetleri ve karma\u015f\u0131kl\u0131klar da vard\u0131r. Uygulaman\u0131z\u0131n sadece do\u011fru \u00e7al\u0131\u015fmas\u0131 de\u011fil, ayn\u0131 zamanda verimli ve \u00f6l\u00e7eklenebilir olmas\u0131 da \u00f6nemlidir. Bu b\u00f6l\u00fcmde, senkronizasyon stratejilerini belirlerken g\u00f6z \u00f6n\u00fcnde bulundurman\u0131z gereken \u00f6nemli noktalar\u0131, olas\u0131 tuzaklar\u0131 ve alternatif yakla\u015f\u0131mlar\u0131 inceleyece\u011fiz.<\/p>\n
Kilitlerin Performans Maliyeti<\/h3>\nHer acquire()<\/code> ve release()<\/code> \u00e7a\u011fr\u0131s\u0131, yorumlay\u0131c\u0131 ve i\u015fletim sistemi \u00fczerinde bir miktar ek y\u00fck olu\u015fturur. Bu, \u00f6zellikle k\u0131sa s\u00fcreli veya \u00e7ok s\u0131k eri\u015filen kritik b\u00f6l\u00fcmlerde belirginle\u015febilir. Kilitler, i\u015f par\u00e7ac\u0131klar\u0131n\u0131n s\u0131raya girmesine ve di\u011ferlerinin bitmesini beklemesine neden olarak e\u015f zamanl\u0131l\u0131\u011f\u0131 azalt\u0131r. Buna \"kilit \u00e7eki\u015fmesi\" (lock contention) denir. \u00c7ok fazla kilit \u00e7eki\u015fmesi, uygulaman\u0131z\u0131n i\u015f par\u00e7ac\u0131\u011f\u0131 say\u0131s\u0131n\u0131 art\u0131rd\u0131k\u00e7a performans\u0131n\u0131n d\u00fc\u015fmesine neden olabilir. \u00c7\u00fcnk\u00fc i\u015f par\u00e7ac\u0131klar\u0131n\u0131n \u00e7o\u011fu, bir kilit \u00fczerinde s\u00fcrekli olarak birbirlerini bekler hale gelirler.<\/p>\n
\n\n      Kilit Kapsam\u0131 (Granularity):<\/strong> Kilitlemenin kapsam\u0131 ne kadar dar olursa, yani kritik b\u00f6l\u00fcm ne kadar k\u0131sa olursa, kilit \u00e7eki\u015fmesi o kadar az olur. T\u00fcm bir fonksiyonu kilitlemek yerine, yaln\u0131zca payla\u015f\u0131ml\u0131 kayna\u011fa eri\u015fen minimum kod blo\u011funu kilitlemek daha verimlidir. Bu \"ince taneli kilitleme\" (fine-grained locking) olarak bilinir. Ancak \u00e7ok ince taneli kilitleme, karma\u015f\u0131kl\u0131\u011f\u0131 art\u0131rabilir ve y\u00f6netimi zorla\u015ft\u0131rabilir. \u0130yi bir denge bulmak \u00f6nemlidir.\n    <\/li>\n
\n      \u00d6l\u00fc Kilitler (Deadlocks):<\/strong> \u0130ki veya daha fazla i\u015f par\u00e7ac\u0131\u011f\u0131, birbirlerinin serbest b\u0131rakmas\u0131n\u0131 bekledi\u011fi kilitleri tuttu\u011funda bir \u00f6l\u00fc kilit durumu ortaya \u00e7\u0131kar. \u00d6rne\u011fin, \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 A, Kilit 1'i tutarken Kilit 2'yi bekler; \u0130\u015f Par\u00e7ac\u0131\u011f\u0131 B, Kilit 2'yi tutarken Kilit 1'i bekler. Bu durumda, her iki i\u015f par\u00e7ac\u0131\u011f\u0131 da s\u00fcresiz olarak ask\u0131da kal\u0131r ve uygulama kilitlenir. \u00d6l\u00fc kilitler, kilitlerin elde edilme s\u0131ras\u0131n\u0131n tutarl\u0131 bir \u015fekilde y\u00f6netilmesiyle \u00f6nlenebilir. Genellikle, t\u00fcm kilitler i\u00e7in global bir s\u0131ralama tan\u0131mlan\u0131r ve i\u015f par\u00e7ac\u0131klar\u0131 her zaman bu s\u0131raya g\u00f6re kilitleri elde etmeye \u00e7al\u0131\u015f\u0131r.\n    <\/li>\n<\/ul>\nmultiprocessing<\/code> vs. threading<\/code>: Ne Zaman Hangisi?<\/h3>\nDaha \u00f6nce bahsetti\u011fimiz Python GIL (Global Interpreter Lock) nedeniyle, Python'da ger\u00e7ek paralellik elde etmek i\u00e7in genellikle multiprocessing<\/code> mod\u00fcl\u00fc tercih edilir. multiprocessing<\/code> mod\u00fcl\u00fc, her biri kendi Python yorumlay\u0131c\u0131s\u0131na ve dolay\u0131s\u0131yla kendi GIL'ine sahip ayr\u0131 i\u015flem (process) olu\u015fturur. Bu, i\u015flemcilerin t\u00fcm \u00e7ekirdeklerinin CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revleri paralel olarak y\u00fcr\u00fctmek i\u00e7in kullan\u0131lmas\u0131n\u0131 sa\u011flar.<\/p>\n\n\n\n\n\n\n\n\n\n\n\u00d6zellik<\/th>\n threading (\u0130\u015f Par\u00e7ac\u0131\u011f\u0131)<\/th>\n multiprocessing (\u0130\u015flem)<\/th>\n<\/tr>\n<\/thead>\n
Paralellik<\/strong><\/td>\n Sahte (GIL nedeniyle CPU-a\u011f\u0131rl\u0131kl\u0131 i\u00e7in)<\/td>\n Ger\u00e7ek (Her i\u015flem kendi yorumlay\u0131c\u0131s\u0131na sahip)<\/td>\n<\/tr>\n
Bellek Payla\u015f\u0131m\u0131<\/strong><\/td>\n Ayn\u0131 bellek alan\u0131n\u0131 payla\u015f\u0131r<\/td>\n Ayr\u0131 bellek alanlar\u0131na sahip<\/td>\n<\/tr>\n
Veri \u0130leti\u015fimi<\/strong><\/td>\n Do\u011frudan (kilitlerle korunmal\u0131)<\/td>\n IPC (Kuyruk, Pipe, Payla\u015f\u0131ml\u0131 Bellek)<\/td>\n<\/tr>\n
Ba\u015flatma Maliyeti<\/strong><\/td>\n Daha d\u00fc\u015f\u00fck (hafif)<\/td>\n Daha y\u00fcksek (a\u011f\u0131r)<\/td>\n<\/tr>\n
En \u0130yi Kullan\u0131m<\/strong><\/td>\n I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler<\/td>\n CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler<\/td>\n<\/tr>\n
Yar\u0131\u015f Durumu<\/strong><\/td>\n Senkronizasyon \u015fart<\/td>\n Daha az riskli (payla\u015f\u0131ml\u0131 bellek hari\u00e7)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\nKarar verirken, uygulaman\u0131z\u0131n darbo\u011faz\u0131n\u0131n nerede oldu\u011funu anlamak kritiktir. E\u011fer uygulaman\u0131z a\u011f\u0131rl\u0131kl\u0131 olarak a\u011f istekleri, dosya i\u015flemleri veya veritaban\u0131 sorgular\u0131 gibi I\/O i\u015flemlerini beklemekle ge\u00e7iyorsa, threading<\/code> genellikle yeterli ve daha hafif bir \u00e7\u00f6z\u00fcm sunar. Ancak yo\u011fun hesaplama gerektiren (matematiksel analiz, bilimsel sim\u00fclasyonlar vb.) CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revleriniz varsa, multiprocessing<\/code> ger\u00e7ek performans art\u0131\u015flar\u0131 sa\u011flayacakt\u0131r. Bu ikisi, Python'da e\u015f zamanl\u0131l\u0131k ve paralellik sorunlar\u0131na y\u00f6nelik farkl\u0131, ancak tamamlay\u0131c\u0131 yakla\u015f\u0131mlard\u0131r.<\/p>\n
Senkronizasyon Stratejileri ve Tasar\u0131m \u0130lkeleri<\/h3>\n\n\n      Payla\u015f\u0131ml\u0131 Durumu Minimumda Tutun:<\/strong> M\u00fcmk\u00fcn oldu\u011funca, i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda payla\u015f\u0131lan veri miktar\u0131n\u0131 azalt\u0131n. Her i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kendi \u00f6zel verileri \u00fczerinde \u00e7al\u0131\u015fmas\u0131, senkronizasyon ihtiyac\u0131n\u0131 ortadan kald\u0131r\u0131r.\n    <\/li>\n
\n      Immutable (De\u011fi\u015fmez) Nesneler Kullan\u0131n:<\/strong> E\u011fer bir nesne olu\u015fturulduktan sonra de\u011fi\u015ftirilemezse, birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 taraf\u0131ndan ayn\u0131 anda okunmas\u0131nda hi\u00e7bir sorun olmaz. Bu, yar\u0131\u015f durumu riskini ortadan kald\u0131r\u0131r.\n    <\/li>\n
\n      Thread-Safe K\u00fct\u00fcphaneleri Kullan\u0131n:<\/strong> queue.Queue<\/code> gibi Python'\u0131n standart k\u00fct\u00fcphanesindeki yerle\u015fik i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli veri yap\u0131lar\u0131n\u0131 ve k\u00fct\u00fcphanelerini tercih edin. Kendi senkronizasyon mekanizmalar\u0131n\u0131z\u0131 yazmaktan ka\u00e7\u0131n\u0131n, \u00e7\u00fcnk\u00fc bu hatalara yol a\u00e7maya daha yatk\u0131nd\u0131r.\n    <\/li>\n
\n      Dikkatli Test Yap\u0131n:<\/strong> \u00c7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalar\u0131n test edilmesi zor olabilir, \u00e7\u00fcnk\u00fc yar\u0131\u015f durumlar\u0131 nadiren ve belirli zamanlama ko\u015fullar\u0131nda ortaya \u00e7\u0131kar. Hata ay\u0131klamay\u0131 kolayla\u015ft\u0131rmak i\u00e7in tekrarlanabilir test senaryolar\u0131 olu\u015fturmak ve hatta zay\u0131f donan\u0131m \u00fczerinde veya farkl\u0131 y\u00fck ko\u015fullar\u0131nda test yapmak \u00f6nemlidir.\n    <\/li>\n<\/ul>\n\n    Uzman \u0130pucu: Karma\u015f\u0131k \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 uygulamalar\u0131nda kilitlenme (deadlock) riskini azaltmak i\u00e7in, t\u00fcm kilitlerinizi tek bir noktada (veya \u00e7ok az say\u0131da iyi tan\u0131mlanm\u0131\u015f noktada) y\u00f6netmeyi ve kilitlerin elde edilme s\u0131ras\u0131n\u0131 her zaman ayn\u0131 tutmay\u0131 d\u00fc\u015f\u00fcn\u00fcn. Ayr\u0131ca, m\u00fcmk\u00fcnse timeout<\/code> parametresiyle acquire()<\/code> metodunu kullanarak s\u00fcresiz beklemeyi \u00f6nleyebilirsiniz.
\n  <\/aside>\nYar\u0131\u015f durumlar\u0131n\u0131 \u00f6nlemek sadece kodunuzun do\u011fru \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flamakla kalmaz, ayn\u0131 zamanda uygulaman\u0131z\u0131n performansl\u0131 ve gelecekteki b\u00fcy\u00fcmelere uyum sa\u011flayabilir olmas\u0131n\u0131 da etkiler. Do\u011fru senkronizasyon ara\u00e7lar\u0131n\u0131 se\u00e7mek, kilitlerin kapsam\u0131n\u0131 dikkatli bir \u015fekilde belirlemek ve multiprocessing<\/code> gibi alternatifleri de\u011ferlendirmek, sa\u011flam ve \u00f6l\u00e7eklenebilir \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 Python uygulamalar\u0131 geli\u015ftirmenin anahtarlar\u0131d\u0131r.<\/p>\n
\u00c7oklu \u0130\u015f Par\u00e7ac\u0131kl\u0131 Uygulamalarda Kullan\u0131c\u0131 Deneyimi ve Mobil Uyum<\/h2>\nPython'da \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 ve senkronizasyon teknikleri genellikle arka u\u00e7 (backend) i\u015fleme, veri y\u00f6netimi ve sunucu tabanl\u0131 uygulamalar\u0131n performans\u0131n\u0131 art\u0131rmak i\u00e7in kullan\u0131l\u0131r. Ancak, bir uygulaman\u0131n son kullan\u0131c\u0131ya sundu\u011fu deneyim, arka u\u00e7ta ger\u00e7ekle\u015fen i\u015flemlerden do\u011frudan etkilenir. Bir web uygulamas\u0131, bir mobil uygulama API'si veya bir masa\u00fcst\u00fc uygulamas\u0131 i\u00e7in geli\u015ftirme yaparken, Python'\u0131n \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 yeteneklerini kullanmak, \u00f6zellikle I\/O-a\u011f\u0131rl\u0131kl\u0131 i\u015flemlerde, kullan\u0131c\u0131 aray\u00fcz\u00fcn\u00fcn (UI) donmadan veya yava\u015flamadan yan\u0131t vermeye devam etmesini sa\u011flayabilir. Sonu\u00e7ta, h\u0131zl\u0131 ve ak\u0131c\u0131 bir kullan\u0131c\u0131 deneyimi, iyi bir arka u\u00e7 performans\u0131yla ba\u015flar.<\/p>\n
\u00d6rne\u011fin, bir kullan\u0131c\u0131n\u0131n b\u00fcy\u00fck bir rapor olu\u015fturma iste\u011fi g\u00f6nderdi\u011fini d\u00fc\u015f\u00fcnelim. Arka u\u00e7taki bir Python uygulamas\u0131, bu raporu bir i\u015f par\u00e7ac\u0131\u011f\u0131nda (veya i\u015flemde) e\u015f zamans\u0131z olarak i\u015flerken, ana i\u015f par\u00e7ac\u0131\u011f\u0131 di\u011fer kullan\u0131c\u0131 isteklerini veya kullan\u0131c\u0131n\u0131n aray\u00fcz\u00fcndeki di\u011fer etkile\u015fimleri i\u015flemeye devam edebilir. Bu, kullan\u0131c\u0131n\u0131n \"uygulama dondu\" hissini ya\u015famas\u0131n\u0131 engeller ve genel kullan\u0131c\u0131 deneyimini iyile\u015ftirir.<\/p>\n
Arka U\u00e7tan \u00d6n Y\u00fcze Veri Ak\u0131\u015f\u0131<\/strong><\/p>\n
\u00c7oklu i\u015f par\u00e7ac\u0131kl\u0131 Python uygulaman\u0131z\u0131n i\u015fledi\u011fi veriler, genellikle bir API arac\u0131l\u0131\u011f\u0131yla veya web soketleri \u00fczerinden \u00f6n y\u00fcze (frontend) iletilir. Bu verilerin h\u0131zl\u0131 ve hatas\u0131z bir \u015fekilde iletilmesi, kullan\u0131c\u0131 aray\u00fcz\u00fcn\u00fcn g\u00fcncel bilgileri dinamik olarak g\u00f6sterebilmesi i\u00e7in kritiktir. \u00d6rne\u011fin, anl\u0131k bildirimler, canl\u0131 durum g\u00fcncellemeleri veya arka planda tamamlanan g\u00f6revlerin sonu\u00e7lar\u0131 gibi bilgiler, \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 sayesinde e\u015f zamanl\u0131 olarak i\u015flenip \u00f6n y\u00fcze g\u00f6nderilebilir. Bu, \u00f6zellikle modern web ve mobil uygulamalarda aranan bir \u00f6zelliktir.<\/p>\n
Mobil Uyumlu Tasar\u0131m\u0131n \u00d6nemi<\/strong><\/p>\n
G\u00fcn\u00fcm\u00fczde kullan\u0131c\u0131lar\u0131n b\u00fcy\u00fck bir \u00e7o\u011funlu\u011fu uygulamalara mobil cihazlar \u00fczerinden eri\u015fti\u011fi i\u00e7in, uygulaman\u0131z\u0131n hem backend'inin h\u0131zl\u0131 olmas\u0131 hem de frontend'inin mobil cihazlara uyumlu olmas\u0131 vazge\u00e7ilmezdir. Python arka ucunuz ne kadar h\u0131zl\u0131 \u00e7al\u0131\u015f\u0131rsa \u00e7al\u0131\u015fs\u0131n, e\u011fer kullan\u0131c\u0131 aray\u00fcz\u00fc farkl\u0131 ekran boyutlar\u0131na ve cihazlara adapte olam\u0131yorsa, kullan\u0131c\u0131 deneyimi k\u00f6t\u00fc olacakt\u0131r. Mobil uyumlu tasar\u0131m (responsive design), web sayfalar\u0131n\u0131n veya uygulamalar\u0131n farkl\u0131 ekran boyutlar\u0131na (telefon, tablet, masa\u00fcst\u00fc) dinamik olarak adapte olmas\u0131n\u0131 sa\u011flayan bir yakla\u015f\u0131md\u0131r.<\/p>\n
Bu adaptasyonu sa\u011flaman\u0131n temel ara\u00e7lar\u0131ndan biri CSS Medya Sorgular\u0131 (Media Queries)'d\u0131r. Medya sorgular\u0131, web taray\u0131c\u0131s\u0131n\u0131n belirli \u00f6zelliklerini (ekran geni\u015fli\u011fi, cihaz tipi vb.) kontrol ederek, farkl\u0131 stil kurallar\u0131n\u0131n uygulanmas\u0131na olanak tan\u0131r. A\u015fa\u011f\u0131da, bir Python arka ucu taraf\u0131ndan sa\u011flanan verilerin, mobil uyumlu bir aray\u00fczde nas\u0131l g\u00f6r\u00fcnt\u00fclenebilece\u011fine dair *kavramsal* bir HTML ve CSS medya sorgular\u0131 \u00f6rne\u011fi bulunmaktad\u0131r. Bu kod do\u011frudan Python \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 mant\u0131\u011f\u0131na ait olmasa da, bir Python backend'inin destekleyece\u011fi bir frontend'in mobil uyumlulu\u011funu g\u00f6stermektedir. Bu t\u00fcr bir yap\u0131, Python'daki multithreading'in sa\u011flad\u0131\u011f\u0131 verileri en iyi \u015fekilde sunman\u0131n bir yolu olarak d\u00fc\u015f\u00fcn\u00fclebilir.<\/p>\n
\n\n

\u00d6zellik<\/th>\n	threading (\u0130\u015f Par\u00e7ac\u0131\u011f\u0131)<\/th>\n	multiprocessing (\u0130\u015flem)<\/th>\n<\/tr>\n<\/thead>\n
Paralellik<\/strong><\/td>\n	Sahte (GIL nedeniyle CPU-a\u011f\u0131rl\u0131kl\u0131 i\u00e7in)<\/td>\n	Ger\u00e7ek (Her i\u015flem kendi yorumlay\u0131c\u0131s\u0131na sahip)<\/td>\n<\/tr>\n
Bellek Payla\u015f\u0131m\u0131<\/strong><\/td>\n	Ayn\u0131 bellek alan\u0131n\u0131 payla\u015f\u0131r<\/td>\n	Ayr\u0131 bellek alanlar\u0131na sahip<\/td>\n<\/tr>\n
Veri \u0130leti\u015fimi<\/strong><\/td>\n	Do\u011frudan (kilitlerle korunmal\u0131)<\/td>\n	IPC (Kuyruk, Pipe, Payla\u015f\u0131ml\u0131 Bellek)<\/td>\n<\/tr>\n
Ba\u015flatma Maliyeti<\/strong><\/td>\n	Daha d\u00fc\u015f\u00fck (hafif)<\/td>\n	Daha y\u00fcksek (a\u011f\u0131r)<\/td>\n<\/tr>\n
En \u0130yi Kullan\u0131m<\/strong><\/td>\n	I\/O-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler<\/td>\n	CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revler<\/td>\n<\/tr>\n
Yar\u0131\u015f Durumu<\/strong><\/td>\n	Senkronizasyon \u015fart<\/td>\n	Daha az riskli (payla\u015f\u0131ml\u0131 bellek hari\u00e7)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Karar verirken, uygulaman\u0131z\u0131n darbo\u011faz\u0131n\u0131n nerede oldu\u011funu anlamak kritiktir. E\u011fer uygulaman\u0131z a\u011f\u0131rl\u0131kl\u0131 olarak a\u011f istekleri, dosya i\u015flemleri veya veritaban\u0131 sorgular\u0131 gibi I\/O i\u015flemlerini beklemekle ge\u00e7iyorsa, threading<\/code> genellikle yeterli ve daha hafif bir \u00e7\u00f6z\u00fcm sunar. Ancak yo\u011fun hesaplama gerektiren (matematiksel analiz, bilimsel sim\u00fclasyonlar vb.) CPU-a\u011f\u0131rl\u0131kl\u0131 g\u00f6revleriniz varsa, multiprocessing<\/code> ger\u00e7ek performans art\u0131\u015flar\u0131 sa\u011flayacakt\u0131r. Bu ikisi, Python'da e\u015f zamanl\u0131l\u0131k ve paralellik sorunlar\u0131na y\u00f6nelik farkl\u0131, ancak tamamlay\u0131c\u0131 yakla\u015f\u0131mlard\u0131r.<\/p>\n Senkronizasyon Stratejileri ve Tasar\u0131m \u0130lkeleri<\/h3>\n\n\n Payla\u015f\u0131ml\u0131 Durumu Minimumda Tutun:<\/strong> M\u00fcmk\u00fcn oldu\u011funca, i\u015f par\u00e7ac\u0131klar\u0131 aras\u0131nda payla\u015f\u0131lan veri miktar\u0131n\u0131 azalt\u0131n. Her i\u015f par\u00e7ac\u0131\u011f\u0131n\u0131n kendi \u00f6zel verileri \u00fczerinde \u00e7al\u0131\u015fmas\u0131, senkronizasyon ihtiyac\u0131n\u0131 ortadan kald\u0131r\u0131r.\n <\/li>\n \n Immutable (De\u011fi\u015fmez) Nesneler Kullan\u0131n:<\/strong> E\u011fer bir nesne olu\u015fturulduktan sonra de\u011fi\u015ftirilemezse, birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 taraf\u0131ndan ayn\u0131 anda okunmas\u0131nda hi\u00e7bir sorun olmaz. Bu, yar\u0131\u015f durumu riskini ortadan kald\u0131r\u0131r.\n <\/li>\n \n Thread-Safe K\u00fct\u00fcphaneleri Kullan\u0131n:<\/strong> queue.Queue<\/code> gibi Python'\u0131n standart k\u00fct\u00fcphanesindeki yerle\u015fik i\u015f par\u00e7ac\u0131\u011f\u0131 a\u00e7\u0131s\u0131ndan g\u00fcvenli veri yap\u0131lar\u0131n\u0131 ve k\u00fct\u00fcphanelerini tercih edin. Kendi senkronizasyon mekanizmalar\u0131n\u0131z\u0131 yazmaktan ka\u00e7\u0131n\u0131n, \u00e7\u00fcnk\u00fc bu hatalara yol a\u00e7maya daha yatk\u0131nd\u0131r.\n <\/li>\n \n Dikkatli Test Yap\u0131n:<\/strong> \u00c7oklu i\u015f par\u00e7ac\u0131kl\u0131 uygulamalar\u0131n test edilmesi zor olabilir, \u00e7\u00fcnk\u00fc yar\u0131\u015f durumlar\u0131 nadiren ve belirli zamanlama ko\u015fullar\u0131nda ortaya \u00e7\u0131kar. Hata ay\u0131klamay\u0131 kolayla\u015ft\u0131rmak i\u00e7in tekrarlanabilir test senaryolar\u0131 olu\u015fturmak ve hatta zay\u0131f donan\u0131m \u00fczerinde veya farkl\u0131 y\u00fck ko\u015fullar\u0131nda test yapmak \u00f6nemlidir.\n <\/li>\n<\/ul>\n\n Uzman \u0130pucu: Karma\u015f\u0131k \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 uygulamalar\u0131nda kilitlenme (deadlock) riskini azaltmak i\u00e7in, t\u00fcm kilitlerinizi tek bir noktada (veya \u00e7ok az say\u0131da iyi tan\u0131mlanm\u0131\u015f noktada) y\u00f6netmeyi ve kilitlerin elde edilme s\u0131ras\u0131n\u0131 her zaman ayn\u0131 tutmay\u0131 d\u00fc\u015f\u00fcn\u00fcn. Ayr\u0131ca, m\u00fcmk\u00fcnse timeout<\/code> parametresiyle acquire()<\/code> metodunu kullanarak s\u00fcresiz beklemeyi \u00f6nleyebilirsiniz. \n <\/aside>\nYar\u0131\u015f durumlar\u0131n\u0131 \u00f6nlemek sadece kodunuzun do\u011fru \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flamakla kalmaz, ayn\u0131 zamanda uygulaman\u0131z\u0131n performansl\u0131 ve gelecekteki b\u00fcy\u00fcmelere uyum sa\u011flayabilir olmas\u0131n\u0131 da etkiler. Do\u011fru senkronizasyon ara\u00e7lar\u0131n\u0131 se\u00e7mek, kilitlerin kapsam\u0131n\u0131 dikkatli bir \u015fekilde belirlemek ve multiprocessing<\/code> gibi alternatifleri de\u011ferlendirmek, sa\u011flam ve \u00f6l\u00e7eklenebilir \u00e7oklu i\u015f par\u00e7ac\u0131kl\u0131 Python uygulamalar\u0131 geli\u015ftirmenin anahtarlar\u0131d\u0131r.<\/p>\n \u00c7oklu \u0130\u015f Par\u00e7ac\u0131kl\u0131 Uygulamalarda Kullan\u0131c\u0131 Deneyimi ve Mobil Uyum<\/h2>\nPython'da \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 ve senkronizasyon teknikleri genellikle arka u\u00e7 (backend) i\u015fleme, veri y\u00f6netimi ve sunucu tabanl\u0131 uygulamalar\u0131n performans\u0131n\u0131 art\u0131rmak i\u00e7in kullan\u0131l\u0131r. Ancak, bir uygulaman\u0131n son kullan\u0131c\u0131ya sundu\u011fu deneyim, arka u\u00e7ta ger\u00e7ekle\u015fen i\u015flemlerden do\u011frudan etkilenir. Bir web uygulamas\u0131, bir mobil uygulama API'si veya bir masa\u00fcst\u00fc uygulamas\u0131 i\u00e7in geli\u015ftirme yaparken, Python'\u0131n \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 yeteneklerini kullanmak, \u00f6zellikle I\/O-a\u011f\u0131rl\u0131kl\u0131 i\u015flemlerde, kullan\u0131c\u0131 aray\u00fcz\u00fcn\u00fcn (UI) donmadan veya yava\u015flamadan yan\u0131t vermeye devam etmesini sa\u011flayabilir. Sonu\u00e7ta, h\u0131zl\u0131 ve ak\u0131c\u0131 bir kullan\u0131c\u0131 deneyimi, iyi bir arka u\u00e7 performans\u0131yla ba\u015flar.<\/p>\n \u00d6rne\u011fin, bir kullan\u0131c\u0131n\u0131n b\u00fcy\u00fck bir rapor olu\u015fturma iste\u011fi g\u00f6nderdi\u011fini d\u00fc\u015f\u00fcnelim. Arka u\u00e7taki bir Python uygulamas\u0131, bu raporu bir i\u015f par\u00e7ac\u0131\u011f\u0131nda (veya i\u015flemde) e\u015f zamans\u0131z olarak i\u015flerken, ana i\u015f par\u00e7ac\u0131\u011f\u0131 di\u011fer kullan\u0131c\u0131 isteklerini veya kullan\u0131c\u0131n\u0131n aray\u00fcz\u00fcndeki di\u011fer etkile\u015fimleri i\u015flemeye devam edebilir. Bu, kullan\u0131c\u0131n\u0131n \"uygulama dondu\" hissini ya\u015famas\u0131n\u0131 engeller ve genel kullan\u0131c\u0131 deneyimini iyile\u015ftirir.<\/p>\n Arka U\u00e7tan \u00d6n Y\u00fcze Veri Ak\u0131\u015f\u0131<\/strong><\/p>\n \u00c7oklu i\u015f par\u00e7ac\u0131kl\u0131 Python uygulaman\u0131z\u0131n i\u015fledi\u011fi veriler, genellikle bir API arac\u0131l\u0131\u011f\u0131yla veya web soketleri \u00fczerinden \u00f6n y\u00fcze (frontend) iletilir. Bu verilerin h\u0131zl\u0131 ve hatas\u0131z bir \u015fekilde iletilmesi, kullan\u0131c\u0131 aray\u00fcz\u00fcn\u00fcn g\u00fcncel bilgileri dinamik olarak g\u00f6sterebilmesi i\u00e7in kritiktir. \u00d6rne\u011fin, anl\u0131k bildirimler, canl\u0131 durum g\u00fcncellemeleri veya arka planda tamamlanan g\u00f6revlerin sonu\u00e7lar\u0131 gibi bilgiler, \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 sayesinde e\u015f zamanl\u0131 olarak i\u015flenip \u00f6n y\u00fcze g\u00f6nderilebilir. Bu, \u00f6zellikle modern web ve mobil uygulamalarda aranan bir \u00f6zelliktir.<\/p>\n Mobil Uyumlu Tasar\u0131m\u0131n \u00d6nemi<\/strong><\/p>\n G\u00fcn\u00fcm\u00fczde kullan\u0131c\u0131lar\u0131n b\u00fcy\u00fck bir \u00e7o\u011funlu\u011fu uygulamalara mobil cihazlar \u00fczerinden eri\u015fti\u011fi i\u00e7in, uygulaman\u0131z\u0131n hem backend'inin h\u0131zl\u0131 olmas\u0131 hem de frontend'inin mobil cihazlara uyumlu olmas\u0131 vazge\u00e7ilmezdir. Python arka ucunuz ne kadar h\u0131zl\u0131 \u00e7al\u0131\u015f\u0131rsa \u00e7al\u0131\u015fs\u0131n, e\u011fer kullan\u0131c\u0131 aray\u00fcz\u00fc farkl\u0131 ekran boyutlar\u0131na ve cihazlara adapte olam\u0131yorsa, kullan\u0131c\u0131 deneyimi k\u00f6t\u00fc olacakt\u0131r. Mobil uyumlu tasar\u0131m (responsive design), web sayfalar\u0131n\u0131n veya uygulamalar\u0131n farkl\u0131 ekran boyutlar\u0131na (telefon, tablet, masa\u00fcst\u00fc) dinamik olarak adapte olmas\u0131n\u0131 sa\u011flayan bir yakla\u015f\u0131md\u0131r.<\/p>\n Bu adaptasyonu sa\u011flaman\u0131n temel ara\u00e7lar\u0131ndan biri CSS Medya Sorgular\u0131 (Media Queries)'d\u0131r. Medya sorgular\u0131, web taray\u0131c\u0131s\u0131n\u0131n belirli \u00f6zelliklerini (ekran geni\u015fli\u011fi, cihaz tipi vb.) kontrol ederek, farkl\u0131 stil kurallar\u0131n\u0131n uygulanmas\u0131na olanak tan\u0131r. A\u015fa\u011f\u0131da, bir Python arka ucu taraf\u0131ndan sa\u011flanan verilerin, mobil uyumlu bir aray\u00fczde nas\u0131l g\u00f6r\u00fcnt\u00fclenebilece\u011fine dair kavramsal bir HTML ve CSS medya sorgular\u0131 \u00f6rne\u011fi bulunmaktad\u0131r. Bu kod do\u011frudan Python \u00e7oklu i\u015f par\u00e7ac\u0131\u011f\u0131 mant\u0131\u011f\u0131na ait olmasa da, bir Python backend'inin destekleyece\u011fi bir frontend'in mobil uyumlulu\u011funu g\u00f6stermektedir. Bu t\u00fcr bir yap\u0131, Python'daki multithreading'in sa\u011flad\u0131\u011f\u0131 verileri en iyi \u015fekilde sunman\u0131n bir yolu olarak d\u00fc\u015f\u00fcn\u00fclebilir.<\/p>\n \n\n