clob字段是oracle专门用于存储超长字符串的字段类型，通常varchar2只能存4000个字符串，超过4000个就存不下去了。
若是直接使用 pandas.read_sql(sql, conn)会报错，那么如何将clob读取到python中呢？ python
1.read方法
import cx_Oracle conn = cx_Oracle.connect("user/pwd@ip/db") cur = conn.cursor() # col 是clob字段 cur.execute("select col from table") pram=[] for i in cur: text = i[0].read() pram.appen(text) cur.close() conn.close()
2.使用DBMS_LOB.SUBSTR模块
方法是先提取前2000个字符串，接着提取2000个字符串….，而后拼接起来。sql
import cx_Oracle import pandas as pd conn = cx_Oracle.connect("user/pwd@ip/db") # 这里只提取前6000个字符串 sql1 = "select DBMS_LOB.SUBSTR(col,2000,1) as col1 as fzss from table" # 1-2000个字符串 sql2 = "select DBMS_LOB.SUBSTR(col,4000,2001) as col2 as fzss from table" # 2001-4000个字符串 sql3 = "select DBMS_LOB.SUBSTR(col,6000,4001) as col3 as fzss from table" # 2001-4000个字符串 # 读取数据 df1 = pd.read_sql(sql1, conn) df2 = pd.read_sql(sql2, conn) df3 = pd.read_sql(sql3, conn) # 有些记录不必定是很长的字符串，结果多是None,须要填充空字符串，不然下面的拼接会出错 df1 = df1.fillna('') df2 = df2.fillna('') df3 = df3.fillna('') # 将它们拼接起来,放在tmp字段上 df1['TMP'] = df1.loc['COL1']+df2.loc['COL2']+df3.loc['COL3']
3.将字符串写入clob字段
方法和将字符串导入varchar2字段是同样的，不须要特殊设置。 好比：oracle
id='123' clob='a'*2**20 # 重复2的20次方次 param=[id, colb] sql = "insert into table (id,colb) values(:1, :2)" cursor.execute(sql, param) conn.commit()