psycopg2.DataError:用于编码"UTF8"的无效字节序列:0xa0

Question

我已经对这个错误进行了大量的谷歌搜索,并将其归结为我正在使用的数据库采用不同的编码.

我正在使用的AIX服务器正在运行

psql 8.2.4

 server_encoding                 | LATIN1          |      | Client Connection Defaults / Locale and Formatting                | Sets the server (database) character set encoding.

我正在使用的Windows 2008 R2服务器正在运行

psql(9.3.4)

CREATE DATABASE postgres
WITH OWNER = postgres
   ENCODING = 'UTF8'
   TABLESPACE = pg_default
   LC_COLLATE = 'English_Australia.1252'
   LC_CTYPE = 'English_Australia.1252'
   CONNECTION LIMIT = -1;

COMMENT ON DATABASE postgres
IS 'default administrative connection database';

现在,当我尝试执行我的下面的python脚本时,我得到了这个错误

Traceback (most recent call last):
File "datamain.py", line 39, in <module>
sys.exit(main())
File "datamain.py", line 33, in main
write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
File "datamain.py", line 21, in write_file_to_table
cur.copy_from(f, table, ",")
psycopg2.DataError: invalid byte sequence for encoding "UTF8": 0xa0
CONTEXT:  COPY cms_jobdef, line 15209    

这是我的剧本

import psycopg2
import StringIO
import sys
import pdb

def connect_db(db, usr, pw, hst, prt):
    conn = psycopg2.connect(database=db, user=usr,
    password=pw, host=hst, port=prt)
    return conn

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    cur = connection.cursor()
    cur.copy_to(f, table, ",")
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor()
    cur.copy_from(f, table, ",")
    f.close()
    cur.close()

def main():
    login = open('login.txt','r')
    con_tctmsv64 = connect_db("x", "y",
    login.readline().strip(),
    "d.domain", "c")
    con_S104838 = connect_db("x", "y", "z", "a", "b")
    try:
        write_table_to_file("cms_jobdef.txt", "cms_jobdef", con_tctmsv64)
        write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
    finally:
        con_tctmsv64.close()
        con_S104838.close()

if __name__ == "__main__":
    sys.exit(main())

删除了一些敏感数据.

所以我不确定如何继续下去.据我所知,copy_expert方法可能有助于导出为UTF8编码.但是因为我从中提取数据的服务器正在运行8.2.4我不认为它支持COPY编码格式.

我认为我最好的尝试是尝试在Windows服务器上重新安装带有LATIN1编码的postgre数据库.当我尝试这样做时,我得到以下错误.

所以我很卡住,任何帮助将不胜感激!

更新我通过将默认本地更改为"C"将Windows上的postgre数据库安装为LATIN1编码.然而,这给了我以下错误,并没有看起来像一个可能成功/正确的方法

我也尝试使用PSQL COPY函数在BINARY中编码文件

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    cur = connection.cursor()
    #cur.copy_to(f, table, ",")
    cur.copy_expert("COPY cms_jobdef TO STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor()
    #cur.copy_from(f, table)
    cur.copy_expert("COPY cms_jobdef FROM STDOUT WITH BINARY", f)
    f.close()
    cur.close()

仍然没有运气我得到同样的错误

DataError: invalid byte sequence for encoding "UTF8": 0xa0
CONTEXT:  COPY cms_jobdef, line 15209, column descript

关于菲尔斯的答案,我尝试过这种方法仍然没有成功.

import psycopg2
import StringIO
import sys
import pdb
import codecs

def connect_db(db, usr, pw, hst, prt):
    conn = psycopg2.connect(database=db, user=usr,
    password=pw, host=hst, port=prt)
    return conn

def write_table_to_file(file, table, connection):
    f = open(file, "w")
    #fx = codecs.EncodedFile(f,"LATIN1", "UTF8")
    cur = connection.cursor()
    cur.execute("SHOW client_encoding;")
    print cur.fetchone()
    cur.copy_to(f, table)
    #cur.copy_expert("COPY cms_jobdef TO STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def write_file_to_table(file, table, connection):
    f = open(file,"r")
    cur = connection.cursor() 
    cur.execute("SET CLIENT_ENCODING TO 'LATIN1';")
    cur.execute("SHOW client_encoding;")
    print cur.fetchone()
    cur.copy_from(f, table)
    #cur.copy_expert("COPY cms_jobdef FROM STDOUT WITH BINARY", f)
    f.close()
    cur.close()

def main(): 
    login = open('login.txt','r')
    con_tctmsv64 = connect_db("x", "y",
    login.readline().strip(),
    "ctmtest1.int.corp.sun", "5436")
    con_S104838 = connect_db("x", "y", "z", "t", "5432")
    try:
        write_table_to_file("cms_jobdef.txt", "cms_jobdef", con_tctmsv64)
        write_file_to_table("cms_jobdef.txt", "cms_jobdef", con_S104838)
    finally:
        con_tctmsv64.close()
        con_S104838.close()

if __name__ == "__main__":
    sys.exit(main())

产量

In [4]: %run datamain.py
('sql_ascii',)
('LATIN1',)

In [5]: 

这成功完成但是当我运行时

select * from cms_jobdef;

新数据库中没有任何内容

我甚至尝试将文件格式从LATIN1转换为UTF8.仍然没有运气

奇怪的是当我通过仅使用postgre COPY函数手动完成此过程时它可以工作.我不知道为什么.再一次,任何帮助将不胜感激.

Answer 1

事实证明,有几个选项可以解决这个问题.

更改Phil建议的客户端编码的选项确实有效.

cur.execute("SET CLIENT_ENCODING TO 'LATIN1';")

另一种选择是即时转换数据.我使用了一个名为codecs的python模块来执行此操作.

f = open(file, "w")
fx = codecs.EncodedFile(f,"LATIN1", "UTF8")
cur = connection.cursor()
cur.execute("SHOW client_encoding;")
print cur.fetchone()
cur.copy_to(fx, table)

关键是

fx = codecs.EncodedFile(f,"LATIN1", "UTF8")

我的主要问题是我没有将更改提交到数据库!傻我:)

Answer 2

我正在从 SQL_ASCII 数据库迁移到 UTF8 数据库，并遇到了同样的问题。基于这个答案，我只是将此语句添加到导入脚本的开头：

set client_encoding to 'latin1'

一切似乎都已正确导入。