dict<\/code>): H\u0131zl\u0131 Anahtar-De\u011fer Aramalar\u0131<\/h3>\nPython s\u00f6zl\u00fckleri, hash tablolar\u0131 kullan\u0131larak implemente edilmi\u015ftir, bu da anahtar-de\u011fer \u00e7iftlerine dayal\u0131 h\u0131zl\u0131 arama, ekleme ve silme i\u015flemleri i\u00e7in onlar\u0131 m\u00fckemmel k\u0131lar. Ortalama durumda, bu i\u015flemler O(1) zaman karma\u015f\u0131kl\u0131\u011f\u0131na sahiptir. Ancak, anahtarlar\u0131n hashlenebilir olmas\u0131 gerekir (genellikle de\u011fi\u015fmez nesneler).<\/p>\n
\nimport timeit\n\nmy_dict = {\"a\": 1, \"b\": 2}\nprint(my_dict[\"a\"]) # Ortalama O(1)\n\n# S\u00f6zl\u00fck olu\u015fturma performans\u0131\ntime_dict_comp = timeit.timeit('{i: i*2 for i in range(1000000)}', number=10)\nprint(f\"S\u00f6zl\u00fck comprehension s\u00fcresi: {time_dict_comp:.4f} saniye\")\n <\/pre>\n<\/code><\/p>\n
K\u00fcmeler (set<\/code>): Benzersiz Elemanlar\u0131n H\u0131zl\u0131 Y\u00f6netimi<\/h3>\nK\u00fcmeler, benzersiz elemanlar\u0131 depolamak i\u00e7in kullan\u0131l\u0131r ve s\u00f6zl\u00fckler gibi hash tablolar\u0131 \u00fczerine kuruludur. Bu, eleman ekleme, silme ve varl\u0131k kontrol\u00fcn\u00fc (in<\/code> operat\u00f6r\u00fc) ortalama O(1) yapar. K\u00fcmeler, \u00f6zellikle b\u00fcy\u00fck koleksiyonlarda yinelenen elemanlar\u0131 h\u0131zl\u0131ca kald\u0131rmak veya k\u00fcme operasyonlar\u0131 (kesi\u015fim, birle\u015fim, fark) ger\u00e7ekle\u015ftirmek i\u00e7in idealdir.<\/p>\n\nimport timeit\n\nlist1 = list(range(100000))\nlist2 = list(range(50000, 150000))\n\n# K\u00fcmelerle kesi\u015fim bulma\ntime_set_intersect = timeit.timeit('''\nset1 = set(list1)\nset2 = set(list2)\ncommon_elements = set1.intersection(set2)\n''', globals=globals(), number=10)\nprint(f\"K\u00fcmelerle kesi\u015fim s\u00fcresi: {time_set_intersect:.4f} saniye\")\n <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnekler, do\u011fru veri yap\u0131s\u0131n\u0131 se\u00e7menin, \u00f6zellikle b\u00fcy\u00fck veri setleriyle \u00e7al\u0131\u015f\u0131rken, uygulaman\u0131z\u0131n performans\u0131n\u0131 nas\u0131l k\u00f6kten etkileyebilece\u011fini g\u00f6zler \u00f6n\u00fcne seriyor. Temel olarak, h\u0131zl\u0131 arama veya benzersizlik gerektiren durumlar i\u00e7in s\u00f6zl\u00fckler ve k\u00fcmeler, listelerden \u00e7ok daha verimlidir.<\/p>\n
Y\u00fcksek Performans \u0130\u00e7in Standart Veri Yap\u0131lar\u0131n\u0131 Etkin Kullanma Y\u00f6ntemleri Nelerdir?<\/h2>\n
Python'\u0131n standart veri yap\u0131lar\u0131, do\u011fru kullan\u0131ld\u0131\u011f\u0131nda bile inan\u0131lmaz derecede g\u00fc\u00e7l\u00fc olabilir. Bu b\u00f6l\u00fcmde, list<\/code>, dict<\/code> ve set<\/code> gibi temel yap\u0131lar\u0131 daha etkin hale getirecek pratik y\u00f6ntemleri ve ger\u00e7ek d\u00fcnya senaryolar\u0131ndaki uygulamalar\u0131n\u0131 ele alaca\u011f\u0131z.<\/p>\nListeler \u0130\u00e7in Optimizasyon Stratejileri<\/h3>\n
Listeler genellikle en \u00e7ok kullan\u0131lan veri yap\u0131s\u0131d\u0131r, ancak performans\u0131 art\u0131rmak i\u00e7in baz\u0131 stratejiler mevcuttur:<\/p>\n
\n- Liste \u00dcretimi (List Comprehensions):<\/strong> B\u00fcy\u00fck listeleri olu\u015ftururken,
append()<\/code> metodunu bir d\u00f6ng\u00fc i\u00e7inde kullanmak yerine liste \u00fcretimlerini tercih edin. Bu yap\u0131, Python'\u0131n yorumlay\u0131c\u0131s\u0131 taraf\u0131ndan C d\u00fczeyinde optimize edildi\u011fi i\u00e7in daha h\u0131zl\u0131d\u0131r.<\/li>\nextend()<\/code> Kullan\u0131m\u0131:<\/strong> Birden fazla eleman\u0131 bir listeye eklerken, append()<\/code> metodunu tek tek \u00e7a\u011f\u0131rmak yerine extend()<\/code> metodunu veya +=<\/code> operat\u00f6r\u00fcn\u00fc kullanmak daha verimlidir. Bu y\u00f6ntemler, belle\u011fi tek seferde tahsis ederek yeniden tahsis maliyetlerini azalt\u0131r.<\/li>\ncollections.deque<\/code> Kullan\u0131m\u0131:<\/strong> E\u011fer listenin ba\u015f\u0131ndan s\u0131k\u00e7a eleman ekleme veya \u00e7\u0131karma i\u015flemi yapman\u0131z gerekiyorsa, standart Python listesi yerine collections.deque<\/code> kullan\u0131n. Listelerde bu i\u015flemler O(n) iken, deque<\/code>'de O(1)'dir, bu da b\u00fcy\u00fck performans fark\u0131 yarat\u0131r.<\/li>\n<\/ul>\nS\u00f6zl\u00fckler ve K\u00fcmeler \u0130\u00e7in \u0130pu\u00e7lar\u0131<\/h3>\n
S\u00f6zl\u00fckler ve k\u00fcmeler zaten hash tabanl\u0131 yap\u0131lar\u0131 sayesinde h\u0131zl\u0131 eri\u015fim sa\u011flarlar, ancak baz\u0131 ek ipu\u00e7lar\u0131 performans\u0131 daha da art\u0131rabilir:<\/p>\n
\n- Hashlenebilir Anahtarlar:<\/strong> S\u00f6zl\u00fck anahtarlar\u0131 ve k\u00fcme elemanlar\u0131 her zaman hashlenebilir (immutable) olmal\u0131d\u0131r. Say\u0131lar, dizeler ve demetler ideal anahtar t\u00fcrleridir. Listeler gibi de\u011fi\u015ftirilebilir nesneler kullan\u0131lamaz.<\/li>\n
- Arama ve \u00dcyelik Kontrol\u00fc:<\/strong> B\u00fcy\u00fck veri setlerinde bir eleman\u0131n varl\u0131\u011f\u0131n\u0131 kontrol etmek i\u00e7in
if item in my_dict<\/code> veya if item in my_set<\/code> kullanmak, bir listeyi d\u00f6ng\u00fcyle taramaktan \u00e7ok daha h\u0131zl\u0131d\u0131r (ortalama O(1) vs O(n)).<\/li>\n- S\u00f6zl\u00fck Birle\u015ftirme:<\/strong> Python 3.9 ve sonras\u0131 i\u00e7in s\u00f6zl\u00fckleri birle\u015ftirmek i\u00e7in
|<\/code> operat\u00f6r\u00fcn\u00fc veya dict.update()<\/code> metodunu kullanmak, d\u00f6ng\u00fc i\u00e7inde tek tek eleman eklemekten daha verimlidir.<\/li>\n<\/ul>\nVaka Analizi: B\u00fcy\u00fck Veri Setlerinde Arama Optimizasyonu<\/h3>\n
Bir e-ticaret platformunun milyonlarca \u00fcr\u00fcn i\u00e7eren envanterinde, kullan\u0131c\u0131lar\u0131n \u00fcr\u00fcn ID'si ile h\u0131zl\u0131ca bilgiye eri\u015fmesi kritik \u00f6neme sahiptir. E\u011fer \u00fcr\u00fcn bilgileri bir liste i\u00e7inde tutulursa ve her arama i\u00e7in liste taran\u0131rsa (O(n)), sistemin tepki s\u00fcresi kabul edilemez derecede artacakt\u0131r. Ancak bu veriyi bir s\u00f6zl\u00fck yap\u0131s\u0131nda tutarak arama performans\u0131n\u0131 radikal bir \u015fekilde art\u0131rabiliriz.<\/p>\n
\nimport time\nimport random\n\nproduct_count = 1_000_000\nproduct_data = []\nfor i in range(product_count):\n product_data.append({'id': i, 'name': f'Product_{i}'})\n\n# S\u00f6zl\u00fc\u011fe dayal\u0131 arama i\u00e7in haz\u0131rl\u0131k: ID'yi anahtar olarak kullanma\nproduct_dict_search = {p['id']: p for p in product_data}\n\nsearch_id = random.randint(0, product_count - 1)\n\n# Liste ile arama (generator kullanarak biraz optimize edilmi\u015f hali)\nstart_time = time.perf_counter()\nfound_product_list = next((p for p in product_data if p['id'] == search_id), None)\nend_time = time.perf_counter()\nprint(f\"Liste ile arama s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\n\n# S\u00f6zl\u00fck ile arama\nstart_time = time.perf_counter()\nfound_product_dict = product_dict_search.get(search_id)\nend_time = time.perf_counter()\nprint(f\"S\u00f6zl\u00fck ile arama s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\n <\/pre>\n<\/code><\/p>\n
Bu analizde, liste ile arama milisaniyeler s\u00fcrerken, s\u00f6zl\u00fck ile arama neredeyse anl\u0131k sonu\u00e7 verir. Bu, do\u011fru veri yap\u0131s\u0131 se\u00e7iminin uygulaman\u0131z\u0131n duyarl\u0131l\u0131\u011f\u0131 \u00fczerindeki kritik etkisini a\u00e7\u0131k\u00e7a g\u00f6stermektedir. B\u00fcy\u00fck veri setleriyle \u00e7al\u0131\u015f\u0131rken, arama ve eri\u015fim performans\u0131 kritik oldu\u011funda s\u00f6zl\u00fckler (veya k\u00fcmeler), listelere k\u0131yasla her zaman tercih edilmelidir.<\/p>\n
\n Uzman \u0130pucu: K\u00fc\u00e7\u00fck metin dizilerini birle\u015ftirirken
+<\/code> operat\u00f6r\u00fcn\u00fc d\u00f6ng\u00fc i\u00e7inde kullanmak yerine str.join()<\/code> metodunu tercih edin. join()<\/code> daha verimlidir \u00e7\u00fcnk\u00fc birle\u015ftirmeyi yapmadan \u00f6nce gerekli belle\u011fi tek seferde tahsis eder.\n <\/div>\nPython'\u0131n Koleksiyon Mod\u00fcl\u00fc: Geli\u015fmi\u015f Veri Yap\u0131lar\u0131yla Performans Nas\u0131l Art\u0131r\u0131l\u0131r?<\/h2>\n
Python'\u0131n standart k\u00fct\u00fcphanesi, collections<\/code> mod\u00fcl\u00fc arac\u0131l\u0131\u011f\u0131yla geli\u015ftiricilere daha karma\u015f\u0131k ve performans odakl\u0131 veri yap\u0131lar\u0131 sunar. Bu yap\u0131lar, belirli senaryolarda temel veri yap\u0131lar\u0131n\u0131n s\u0131n\u0131rl\u0131l\u0131klar\u0131n\u0131 a\u015farak \u00f6nemli performans kazan\u0131mlar\u0131 sa\u011flayabilir. Bu b\u00f6l\u00fcmde, collections<\/code> mod\u00fcl\u00fcn\u00fcn en kullan\u0131\u015fl\u0131 bile\u015fenlerini ve bunlar\u0131n nas\u0131l optimize edici bir etki yaratt\u0131\u011f\u0131n\u0131 inceleyece\u011fiz.<\/p>\ncollections.deque<\/code>: \u00c7ift U\u00e7lu Kuyruklarla H\u0131zl\u0131 Ekleme\/\u00c7\u0131karma<\/h3>\ndeque<\/code> (double-ended queue), bir listenin hem ba\u015f\u0131ndan hem de sonundan h\u0131zl\u0131 bir \u015fekilde eleman ekleyip \u00e7\u0131karmak i\u00e7in tasarlanm\u0131\u015ft\u0131r. Standart Python listelerinde listenin ba\u015f\u0131na eleman eklemek veya \u00e7\u0131karmak O(n) iken, deque<\/code> i\u00e7in bu i\u015flemler O(1)'dir. Bu \u00f6zellik, \u00f6zellikle kuyruk veya y\u0131\u011f\u0131n benzeri veri eri\u015fim desenlerine sahip uygulamalarda (\u00f6rn. son i\u015flem ge\u00e7mi\u015fi, mesaj kuyruklar\u0131) \u00e7ok faydal\u0131d\u0131r.<\/p>\n\nfrom collections import deque\nimport timeit\n\nnum_elements = 100000\n\n# Liste ile ba\u015fa ekleme\ntime_list_insert_front = timeit.timeit('''\nmy_list = []\nfor i in range(num_elements):\n my_list.insert(0, i)\n''', globals=globals(), number=1)\nprint(f\"Liste ile ba\u015fa ekleme s\u00fcresi: {time_list_insert_front:.4f} saniye\")\n\n# Deque ile ba\u015fa ekleme\ntime_deque_appendleft = timeit.timeit('''\nmy_deque = deque()\nfor i in range(num_elements):\n my_deque.appendleft(i)\n''', globals=globals(), number=1)\nprint(f\"Deque ile ba\u015fa ekleme s\u00fcresi: {time_deque_appendleft:.4f} saniye\")\n <\/pre>\n<\/code><\/p>\n
Yukar\u0131daki \u00f6rnekte g\u00f6r\u00fcld\u00fc\u011f\u00fc gibi, deque<\/code>'in performans\u0131, listenin ba\u015f\u0131na s\u00fcrekli eleman ekleme senaryolar\u0131nda \u00fcst\u00fcnd\u00fcr.<\/p>\ncollections.namedtuple<\/code>: Hafif, De\u011fi\u015fmez Nesneler<\/h3>\nnamedtuple<\/code>, standart demetleri geni\u015fleterek, her bir elemana isimle eri\u015fim imkan\u0131 sunan, ancak s\u0131radan bir Python s\u0131n\u0131f\u0131n\u0131n bellek y\u00fck\u00fcne sahip olmayan hafif, de\u011fi\u015fmez nesneler olu\u015fturman\u0131z\u0131 sa\u011flar. Bu, \u00f6zellikle veritaban\u0131 kay\u0131tlar\u0131 veya CSV sat\u0131rlar\u0131 gibi yap\u0131land\u0131r\u0131lm\u0131\u015f verileri depolarken hem okunabilirli\u011fi art\u0131r\u0131r hem de bellek verimlili\u011fi sa\u011flar.<\/p>\n\nfrom collections import namedtuple\nimport sys\n\nPoint = namedtuple('Point', ['x', 'y'])\np = Point(10, 20)\nprint(f\"NamedTuple bellek boyutu: {sys.getsizeof(p)} byte\")\n\nclass RegularPoint:\n def __init__(self, x, y):\n self.x = x\n self.y = y\nregular_p = RegularPoint(10, 20)\nprint(f\"Normal s\u0131n\u0131f bellek boyutu: {sys.getsizeof(regular_p)} byte\")\n <\/pre>\n<\/code><\/p>\n
collections.defaultdict<\/code>: Otomatik Varsay\u0131lan De\u011ferli S\u00f6zl\u00fckler<\/h3>\ndefaultdict<\/code>, eri\u015filmeye \u00e7al\u0131\u015f\u0131lan bir anahtar s\u00f6zl\u00fckte yoksa otomatik olarak varsay\u0131lan bir de\u011fer \u00fcreten \u00f6zel bir s\u00f6zl\u00fck t\u00fcr\u00fcd\u00fcr. Bu, \u00f6zellikle saya\u00e7lar olu\u015ftururken veya verileri grupland\u0131r\u0131rken, anahtar\u0131n varl\u0131\u011f\u0131n\u0131 kontrol etme ihtiyac\u0131n\u0131 ortadan kald\u0131r\u0131r ve kodu daha temiz ve verimli hale getirir.<\/p>\n\nfrom collections import defaultdict\n\n# Kelimelerin frekans\u0131n\u0131 sayma\nword_counts = defaultdict(int)\nsentence = \"python is great python is fun\"\nfor word in sentence.split():\n word_counts[word] += 1\nprint(word_counts)\n <\/pre>\n<\/code><\/p>\n
collections.Counter<\/code>: Verimli Say\u0131m \u0130\u015flemleri<\/h3>\nCounter<\/code>, bir iterable i\u00e7indeki elemanlar\u0131n frekans\u0131n\u0131 saymak i\u00e7in \u00f6zel olarak tasarlanm\u0131\u015f bir s\u00f6zl\u00fck alt s\u0131n\u0131f\u0131d\u0131r. Metin analizi, log i\u015fleme veya istatistiksel analiz gibi say\u0131m i\u015flemleri gerektiren her senaryoda inan\u0131lmaz derecede kullan\u0131\u015fl\u0131d\u0131r ve manuel bir s\u00f6zl\u00fck uygulamas\u0131na g\u00f6re \u00e7ok daha h\u0131zl\u0131d\u0131r.<\/p>\n\nfrom collections import Counter\n\ntext = \"bu bir deneme metnidir deneme metni\"\nword_freq = Counter(text.split())\nprint(word_freq.most_common(2)) # En s\u0131k ge\u00e7en 2 kelime\n <\/pre>\n<\/code><\/p>\n
Vaka Analizi: Anl\u0131k Log Analizi ve H\u0131z Optimizasyonu<\/h3>\n
Bir web sunucusunun akan log dosyas\u0131n\u0131 anl\u0131k olarak analiz etti\u011finizi varsayal\u0131m. En s\u0131k hangi IP adreslerinin istek g\u00f6nderdi\u011fini veya hangi hata kodlar\u0131n\u0131n en \u00e7ok tekrar etti\u011fini bulmak istiyorsunuz. Counter<\/code> mod\u00fcl\u00fc, bu t\u00fcr say\u0131m i\u015flemlerini manuel bir s\u00f6zl\u00fck uygulamas\u0131na k\u0131yasla \u00e7ok daha h\u0131zl\u0131 ve verimli bir \u015fekilde ger\u00e7ekle\u015ftirmenizi sa\u011flar.<\/p>\n\nimport random\nimport time\n\nip_addresses = [f\"192.168.1.{i}\" for i in range(1, 10)]\nlog_entries = []\nfor _ in range(100000):\n log_entries.append({'ip': random.choice(ip_addresses)})\n\nstart_time = time.perf_counter()\nip_counts = Counter(entry['ip'] for entry in log_entries)\nend_time = time.perf_counter()\nprint(f\"Counter ile IP sayma s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\nprint(\"En s\u0131k g\u00f6r\u00fclen IP'ler:\", ip_counts.most_common(3))\n <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnek, collections<\/code> mod\u00fcl\u00fcndeki \u00f6zel veri yap\u0131lar\u0131n\u0131n, belirli g\u00f6revler i\u00e7in temel yap\u0131lardan \u00e7ok daha verimli ve okunabilir \u00e7\u00f6z\u00fcmler sunarak Python uygulamalar\u0131n\u0131z\u0131n genel performans\u0131n\u0131 nas\u0131l art\u0131rabilece\u011fini g\u00f6stermektedir.<\/p>\nNumPy ve Pandas ile Bilimsel Hesaplamalarda ve B\u00fcy\u00fck Veri \u0130\u015flemede Performans S\u0131rlar\u0131 Nelerdir?<\/h2>\n
Veri bilimi, makine \u00f6\u011frenimi ve bilimsel hesaplama alanlar\u0131nda Python, NumPy ve Pandas gibi k\u00fct\u00fcphaneler sayesinde bir end\u00fcstri standard\u0131 haline gelmi\u015ftir. Bu k\u00fct\u00fcphaneler, Python'\u0131n yerel veri yap\u0131lar\u0131n\u0131n aksine, C veya Fortran gibi dillerde optimize edilmi\u015f alt seviye implementasyonlar kullanarak devasa veri setlerinde bile ola\u011fan\u00fcst\u00fc performans sunarlar.<\/p>\n
NumPy ndarray<\/code>: Homojen ve Vekt\u00f6rel \u0130\u015flemlerin G\u00fcc\u00fc<\/h3>\nNumPy (Numerical Python), \u00e7ok boyutlu diziler (ndarray<\/code>) i\u00e7in temel bir pakettir. Python listelerinin aksine, NumPy dizileri homojendir; yani t\u00fcm elemanlar\u0131 ayn\u0131 veri tipine sahiptir. Bu \u00f6zellik, bellek d\u00fczenini optimize eder ve C dilindeki dizilere benzer biti\u015fik bellek bloklar\u0131 kullan\u0131lmas\u0131na olanak tan\u0131r. Sonu\u00e7 olarak, elemanlara eri\u015fim ve matematiksel i\u015flemler \u00e7ok daha h\u0131zl\u0131d\u0131r. NumPy, b\u00fcy\u00fck diziler \u00fczerinde \"vekt\u00f6rle\u015ftirilmi\u015f\" operasyonlar yapma yetene\u011fi sayesinde d\u00f6ng\u00fc yazma ihtiyac\u0131n\u0131 ortadan kald\u0131r\u0131r ve CPU'nun SIMD (Single Instruction, Multiple Data) yeteneklerinden faydalan\u0131r.<\/p>\n\nimport numpy as np\nimport timeit\n\npython_list_a = list(range(1000000))\npython_list_b = list(range(1000000))\n\n# Python listesi ile toplama\nstart_time = timeit.default_timer()\nresult_list = [python_list_a[i] + python_list_b[i] for i in range(1000000)]\nend_time = timeit.default_timer()\nprint(f\"Python listesi ile toplama s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\n\n# NumPy dizisi ile toplama (vekt\u00f6rle\u015ftirilmi\u015f)\nnumpy_array_a = np.arange(1000000)\nnumpy_array_b = np.arange(1000000)\n\nstart_time = timeit.default_timer()\nresult_array = numpy_array_a + numpy_array_b\nend_time = timeit.default_timer()\nprint(f\"NumPy dizisi ile toplama s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\n <\/pre>\n<\/code><\/p>\n
Bu \u00f6rnek, NumPy'\u0131n vekt\u00f6rle\u015ftirilmi\u015f i\u015flemlerinin performans \u00fcst\u00fcnl\u00fc\u011f\u00fcn\u00fc \u00e7arp\u0131c\u0131 bir \u015fekilde g\u00f6stermektedir.<\/p>\n
Pandas Series<\/code> ve DataFrame<\/code>: Yap\u0131land\u0131r\u0131lm\u0131\u015f Veride Y\u00fcksek Performans<\/h3>\nPandas, veri manip\u00fclasyonu ve analizi i\u00e7in g\u00fc\u00e7l\u00fc bir k\u00fct\u00fcphanedir. NumPy'\u0131n \u00fczerine in\u015fa edilmi\u015ftir ve iki temel veri yap\u0131s\u0131 sunar: Series<\/code> (tek boyutlu, etiketli diziler) ve DataFrame<\/code> (iki boyutlu, etiketli, tablo benzeri yap\u0131). Pandas, t\u00fcm veri i\u015fleme operasyonlar\u0131n\u0131 m\u00fcmk\u00fcn oldu\u011funca vekt\u00f6rle\u015ftirerek ve arka planda NumPy'dan faydalanarak y\u00fcksek performans sa\u011flar. Bu, b\u00fcy\u00fck CSV dosyalar\u0131n\u0131 okuma, veri temizleme, filtreleme, grupland\u0131rma ve birle\u015ftirme gibi i\u015flemleri Python listeleri veya s\u00f6zl\u00fckleriyle yapmaktan \u00e7ok daha h\u0131zl\u0131 hale getirir.<\/p>\n\nimport pandas as pd\nimport random\n\ndata = {\n 'city': [random.choice(['Istanbul', 'Ankara', 'Izmir']) for _ in range(100000)],\n 'population': [random.randint(50000, 5000000) for _ in range(100000)]\n}\ndf = pd.DataFrame(data)\n\n# \u015eehirlere g\u00f6re n\u00fcfus ortalamas\u0131 hesaplama\nstart_time = timeit.default_timer()\navg_population_by_city = df.groupby('city')['population'].mean()\nend_time = timeit.default_timer()\nprint(f\"\\nPandas ile gruplama ve ortalama alma s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\nprint(avg_population_by_city)\n <\/pre>\n<\/code><\/p>\n
Pandas, b\u00fcy\u00fck veri analizi projelerinde apply<\/code> fonksiyonundan m\u00fcmk\u00fcn oldu\u011funca ka\u00e7\u0131n\u0131p vekt\u00f6rle\u015ftirilmi\u015f i\u015flemleri tercih etmekle performans\u0131 en \u00fcst d\u00fczeye \u00e7\u0131kar\u0131r.<\/p>\nVaka Analizi: Finansal Veri Analizi ve Makine \u00d6\u011frenimi Haz\u0131rl\u0131\u011f\u0131<\/h3>\n
Milyonlarca sat\u0131r hisse senedi fiyat\u0131 verisini analiz etti\u011finizi varsayal\u0131m. Bu veriler \u00fczerinde hareketli ortalamalar, standart sapmalar hesaplaman\u0131z ve ard\u0131ndan bir makine \u00f6\u011frenimi modeline beslemek i\u00e7in \u00f6zellikler t\u00fcretmeniz gerekiyor. NumPy ve Pandas olmadan bu, hem \u00e7ok yava\u015f hem de karma\u015f\u0131k bir g\u00f6rev olacakt\u0131r. Bu k\u00fct\u00fcphaneler, karma\u015f\u0131k operasyonlar\u0131 saniyeler i\u00e7inde tamamlar.<\/p>\n
\nimport pandas as pd\nimport numpy as np\n\ndates = pd.date_range(start='2000-01-01', periods=100000, freq='D')\nprices = np.random.randn(100000).cumsum() + 100\nfinancial_data = pd.DataFrame({'Date': dates, 'Price': prices}).set_index('Date')\n\n# 30 g\u00fcnl\u00fck hareketli ortalama hesaplama\nstart_time = time.perf_counter()\nfinancial_data['MA_30'] = financial_data['Price'].rolling(window=30).mean()\nend_time = time.perf_counter()\nprint(f\"\\n30 g\u00fcnl\u00fck hareketli ortalama hesaplama s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\nprint(\"\\nFinansal veri \u00f6rnekleri:\")\nprint(financial_data.head())\n <\/pre>\n<\/code><\/p>\n
NumPy ve Pandas'\u0131n sundu\u011fu y\u00fcksek performansl\u0131 veri yap\u0131lar\u0131, modern veri odakl\u0131 uygulamalar\u0131n bel kemi\u011fidir.<\/p>\n
Geli\u015fmi\u015f Performans \u0130pu\u00e7lar\u0131 ve En \u0130yi Uygulamalar Nelerdir?<\/h2>\n
K\u0131demli bir geli\u015ftirici olarak, sadece do\u011fru veri yap\u0131lar\u0131n\u0131 se\u00e7mekle kalmay\u0131p, ayn\u0131 zamanda Python'\u0131n performans\u0131n\u0131 daha da ileri ta\u015f\u0131yacak geli\u015fmi\u015f teknikleri ve en iyi uygulamalar\u0131 da bilmeniz gerekir. Bu b\u00f6l\u00fcmde, bellek optimizasyonundan h\u0131zland\u0131rma k\u00fct\u00fcphanelerine ve kodunuzu mobil uyumlu hale getirmeye kadar \u00e7e\u015fitli ipu\u00e7lar\u0131n\u0131 ele alaca\u011f\u0131z.<\/p>\n
Bellek Optimizasyonu \u0130\u00e7in __slots__<\/code> Kullan\u0131m\u0131<\/h3>\nPython'da her s\u0131n\u0131f \u00f6rne\u011fi, varsay\u0131lan olarak bir __dict__<\/code> \u00f6zniteli\u011fine sahiptir. Bu \u00f6znitelik, o \u00f6rne\u011fe ait t\u00fcm de\u011fi\u015fkenleri bir s\u00f6zl\u00fck olarak depolar ve esneklik sa\u011flasa da \u00f6nemli bir bellek y\u00fck\u00fc olu\u015fturur. \u00d6zellikle \u00e7ok say\u0131da nesne \u00f6rne\u011fi olu\u015fturman\u0131z gerekti\u011finde (\u00f6rn. binlerce karakter nesnesi), bu bellek t\u00fcketimi performans\u0131 olumsuz etkileyebilir.<\/p>\n__slots__<\/code> kullanarak, bir s\u0131n\u0131f\u0131n belirli \u00f6zniteliklere sahip olaca\u011f\u0131n\u0131 Python yorumlay\u0131c\u0131s\u0131na \u00f6nceden bildirebilirsiniz. Bu, Python'\u0131n \u00f6znitelikleri do\u011frudan nesnenin kendi belle\u011finde saklamas\u0131n\u0131 sa\u011flar, __dict__<\/code> kullan\u0131m\u0131ndan ka\u00e7\u0131n\u0131r ve b\u00f6ylece bellek t\u00fcketimini azalt\u0131rken \u00f6znitelik eri\u015fimini h\u0131zland\u0131r\u0131r.<\/p>\n\nimport sys\n\nclass PointWithSlots:\n __slots__ = ['x', 'y']\n def __init__(self, x, y):\n self.x = x\n self.y = y\n\nclass PointWithDict: # Varsay\u0131lan davran\u0131\u015f\n def __init__(self, x, y):\n self.x = x\n self.y = y\n\np_slots = PointWithSlots(10, 20)\np_dict = PointWithDict(10, 20)\n\nprint(f\"PointWithSlots bellek boyutu: {sys.getsizeof(p_slots)} byte\")\nprint(f\"PointWithDict bellek boyutu: {sys.getsizeof(p_dict)} byte\")\n <\/pre>\n<\/code><\/p>\n
G\u00f6rd\u00fc\u011f\u00fcn\u00fcz gibi, __slots__<\/code> kullanan nesne \u00f6nemli \u00f6l\u00e7\u00fcde daha az bellek t\u00fcketir. Ancak dinamik \u00f6znitelik ekleme yetene\u011fini kaybedersiniz.<\/p>\nJIT Derleyicileri ile Performans H\u0131zland\u0131rma: Numba ve Cython<\/h3>\n
Yo\u011fun hesaplama gerektiren kod bloklar\u0131 i\u00e7in Python'\u0131n yorumlay\u0131c\u0131 do\u011fas\u0131ndan kaynaklanan performans\u0131 a\u015fmak amac\u0131yla Just-In-Time (JIT) derleyicilerinden faydalanabilirsiniz:<\/p>\n
\n- Numba:<\/strong> NumPy dizileri ve Python fonksiyonlar\u0131 \u00fczerinde \u00e7al\u0131\u015fan bilimsel ve say\u0131sal Python kodunu h\u0131zland\u0131rmak i\u00e7in tasarlanm\u0131\u015f bir JIT derleyicisidir. Kodunuzu do\u011frudan makine koduna derleyerek Python d\u00f6ng\u00fclerinin performans\u0131n\u0131 C h\u0131z\u0131na yakla\u015ft\u0131rabilir.<\/li>\n
\nfrom numba import jit\nimport numpy as np\nimport time\n\n@jit(nopython=True)\ndef fast_sum(arr):\n total = 0.0\n for x in arr:\n total += x\n return total\n\narr = np.random.rand(10**7)\n\nstart_time = time.perf_counter()\nresult = fast_sum(arr) # \u0130lk \u00e7a\u011fr\u0131da derleme yap\u0131l\u0131r\nend_time = time.perf_counter()\nprint(f\"\\nNumba ile toplam s\u00fcresi: {(end_time - start_time) * 1000:.4f} ms\")\n <\/pre>\n<\/code><\/p>\n
- Cython:<\/strong> Python kodunuzu C koduna derleyerek performans art\u0131\u015f\u0131 sa\u011flar. Python s\u00f6zdizimini kullanmaya devam edebilir, iste\u011fe ba\u011fl\u0131 olarak C veri tiplerini de ekleyerek \u00e7ok daha y\u00fcksek h\u0131zlara ula\u015fabilirsiniz.<\/li>\n<\/ul>\n
Kod Profilleme ve Bellek Kullan\u0131m\u0131 Analizi<\/h3>\n
Performans sorunlar\u0131n\u0131 \u00e7\u00f6zmenin ilk ad\u0131m\u0131, nerede olduklar\u0131n\u0131 bilmektir.<\/p>\n
\ntimeit<\/code> Mod\u00fcl\u00fc:<\/strong> K\u00fc\u00e7\u00fck kod par\u00e7ac\u0131klar\u0131n\u0131n \u00e7al\u0131\u015fma s\u00fcresini \u00f6l\u00e7mek i\u00e7in kullan\u0131l\u0131r.<\/li>\ncProfile<\/code> Mod\u00fcl\u00fc:<\/strong> Uygulaman\u0131z\u0131n her fonksiyon \u00e7a\u011fr\u0131s\u0131n\u0131 analiz ederek hangi fonksiyonlar\u0131n ne kadar zaman harcad\u0131\u011f\u0131n\u0131 g\u00f6sterir.<\/li>\nmemory_profiler<\/code> K\u00fct\u00fcphanesi:<\/strong> Kodunuzun farkl\u0131 noktalar\u0131nda bellek t\u00fcketimini sat\u0131r sat\u0131r analiz etmenize olanak tan\u0131r.<\/li>\n<\/ul>\nMobil Uyumluluk \u0130\u00e7in HTML ve CSS Yakla\u015f\u0131mlar\u0131<\/h3>\n
Geli\u015ftirdi\u011finiz Python uygulamalar\u0131n\u0131n (\u00f6zellikle web tabanl\u0131 raporlar veya g\u00f6rselle\u015ftirmeler) \u00e7\u0131kt\u0131lar\u0131n\u0131n farkl\u0131 cihazlarda d\u00fczg\u00fcn g\u00f6r\u00fcnt\u00fclenmesi, kullan\u0131c\u0131 deneyimi a\u00e7\u0131s\u0131ndan \u00f6nemlidir. \u00d6zellikle veri tablolar\u0131 veya interaktif grafikler gibi kompleks HTML \u00f6\u011feleri i\u00e7eren Python web uygulamalar\u0131nda mobil uyumluluk d\u00fc\u015f\u00fcn\u00fclmelidir. CSS media query'ler arac\u0131l\u0131\u011f\u0131yla ekran boyutuna g\u00f6re tablolar\u0131n d\u00fczenini de\u011fi\u015ftirebilirsiniz. \u0130\u015fte basit bir \u00f6rnek:<\/p>\n
\n