使用 Python 将 2Gb+ SELECT 导出为 CSV（内存不足）

Question

我正在尝试从 Netezza 导出一个大文件（使用 Netezza ODBC + pyodbc），此解决方案会引发 memoryError，如果我在没有“list”的情况下循环，它会非常慢。你知道一个不会杀死我的服务器/python进程但可以运行得更快的中间解决方案吗？

cursorNZ.execute(sql)
archi = open("c:\test.csv", "w")
lista = list(cursorNZ.fetchall())
for fila  in lista:
    registro = ''
    for campo in fila:
        campo = str(campo)
        registro = registro+str(campo)+";"
    registro = registro[:-1]
    registro = registro.replace('None','NULL')
    registro = registro.replace("'NULL'","NULL")
    archi.write(registro+"\n")

- - 编辑 - -

谢谢，我正在尝试：其中“sql”是查询，cursorNZ 是

connMy = pyodbc.connect(DRIVER=.....)
cursorNZ = connNZ.cursor()

chunk = 10 ** 5  # tweak this
chunks = pandas.read_sql(sql, cursorNZ, chunksize=chunk)
with open('C:/test.csv', 'a') as output:
    for n, df in enumerate(chunks):
        write_header = n == 0
        df.to_csv(output, sep=';', header=write_header, na_rep='NULL')

有这个： AttributeError: 'pyodbc.Cursor' object has no attribute 'cursor' 知道吗？

Answer 1

不要使用cursorNZ.fetchall().

相反，直接循环光标：

with open("c:/test.csv", "w") as archi:  # note the fixed '/'
    cursorNZ.execute(sql)
    for fila in cursorNZ:
        registro = ''
        for campo in fila:
            campo = str(campo)
            registro = registro+str(campo)+";"
        registro = registro[:-1]
        registro = registro.replace('None','NULL')
        registro = registro.replace("'NULL'","NULL")
        archi.write(registro+"\n")

就个人而言，我只会使用熊猫：

import pyodbc
import pandas

cnn = pyodbc.connect(DRIVER=.....)
chunksize = 10 ** 5  # tweak this
chunks = pandas.read_sql(sql, cnn, chunksize=chunksize)

with open('C:/test.csv', 'a') as output:
    for n, df in enumerate(chunks):
        write_header = n == 0
        df.to_csv(output, sep=';', header=write_header, na_rep='NULL')