使用 Psycopg2 和 unnest 時的“未知”數據類型

Question

我正在嘗試使用 python 將 UK Companies House csv 文件中的數據批量加載到 PostgreSQL 中。

我將每一行數據轉換成一個字典列表，然后使用 unnest 語句將數據解包成一個單一的高容量 sql 語句，這是我正在做的一個示例，（源中有更多的字段）...

def buildDict(row)
    clean_name = row[0].decode('utf-8').upper()
    country_code = lookups.getCountryCodeFromName(row[14])
    if len(country_code) > 2:
        country_code = None
        insert_dict = {
            'companyname': row[0],
            'companynumber': row[1],
            'regaddress_careof': row[2],
            'regaddress_pobox': row[3],
            'dissolutiondate': row[13],
            }

            # convert 'None' and '' strings to None
            for k, v in six.iteritems(insert_dict):
                insert_dict[k] = set_to_null(v)


def fastInsert(data):
    sql='''
        INSERT INTO uk_data.companies_house(
          companyname,
          companynumber,
          regaddress_careof,
          regaddress_pobox,
          dissolutiondate
          )
          SELECT
          unnest( %(companyname)s ),
          unnest( %(companynumber)s ),
          unnest( %(regaddress_careof)s ),
          unnest( %(regaddress_pobox)s ),
          unnest( %(dissolutiondate)s )
          ;
    '''

    companyname=[str(r['companyname']) for r in data]
    companynumber=[str(r['companynumber']) for r in data]
    regaddress_careof=[str(r['regaddress_careof']) for r in data]
    regaddress_pobox=[str(r['regaddress_pobox']) for r in data]
    dissolutiondate=[datetime.strptime(r['dissolutiondate'], "%d/%m/%Y") if r['dissolutiondate'] else None for r in data]
    execute(sql,locals())


def execute(sql,params={}):
    with connect() as connection:
        with connection.cursor() as cursor:
            if params:
                cursor.execute(sql,params)
            else:
                cursor.execute(sql)

只要將所有內容都轉換為字符串，此代碼就可以正常工作，但是當我嘗試將數據轉換為日期時，每次日期記錄沒有值時都會出現以下錯誤（注意，此值被條件設置為None ，所以應該加載到 PostgreSQL 中）。

Error could not determine polymorphic type because input has type "unknown"

我嘗試在 unnest 語句中將類型轉換為::DATE ，如下所示：

sql='''
    INSERT INTO uk_data.companies_house(
      companyname,
      companynumber,
      regaddress_careof,
      regaddress_pobox,
      dissolutiondate
      )
      SELECT
      unnest( %(companyname)s ),
      unnest( %(companynumber)s ),
      unnest( %(regaddress_careof)s ),
      unnest( %(regaddress_pobox)s ),
      unnest( %(dissolutiondate)s )::DATE
      ;
'''

但這無濟於事。 我的當地人的打印顯示單個記錄的以下內容：

('these are my locals: ', {'regaddress_posttown': ['LEEDS'], 'regaddress_addressline1': ['METROHOUSE 57 PEPPER ROAD'], 'regaddress_addressline2': ['HUNSLET'], 'regaddress_careof': ['None'], 'companystatus': ['Active'], 'companycategory': ['Private Limited Company'], 'companyname': ['! LTD'], 'countryoforigin': ['None'], 'regaddress_pobox': ['None'], 'regaddress_country': ['None'], 'dissolutiondate': None, 'regaddress_postcode': ['LS10 2RU'], 'regaddress_county': ['YORKSHIRE'], 'sql': '
        INSERT INTO uk_data.companies_house(
          companyname,
          companynumber,
          regaddress_careof,
          regaddress_pobox,
          regaddress_addressline1,
          regaddress_addressline2,
          regaddress_posttown,
          regaddress_county,
          regaddress_country,
          regaddress_postcode,
          companycategory,
          companystatus,
          countryoforigin,
          dissolutiondate
          )
          SELECT
          unnest( %(companyname)s ),
          unnest( %(companynumber)s ),
          unnest( %(regaddress_careof)s ),
          unnest( %(regaddress_pobox)s ),
          unnest( %(regaddress_addressline1)s ),
          unnest( %(regaddress_addressline2)s ),
          unnest( %(regaddress_posttown)s ),
          unnest( %(regaddress_county)s ),
          unnest( %(regaddress_country)s ),
          unnest( %(regaddress_postcode)s ),
          unnest( %(companycategory)s ),
          unnest( %(companystatus)s ),
          unnest( %(countryoforigin)s ),
          unnest( %(dissolutiondate)s )
          ;
    ', 'r': {'regaddress_posttown': 'LEEDS', 'regaddress_careof': None, 'companystatus': 'Active', 'companynumber': '08209948', 'regaddress_addressline1': 'METROHOUSE 57 PEPPER ROAD', 'regaddress_addressline2': 'HUNSLET', 'companycategory': 'Private Limited Company', 'companyname': '! LTD', 'countryoforigin': None, 'regaddress_pobox': None, 'regaddress_country': None, 'dissolutiondate': None, 'regaddress_postcode': 'LS10 2RU', 'regaddress_county': 'YORKSHIRE'}, 'data': [{'regaddress_posttown': 'LEEDS', 'regaddress_careof': None, 'companystatus': 'Active', 'companynumber': '08209948', 'regaddress_addressline1': 'METROHOUSE 57 PEPPER ROAD', 'regaddress_addressline2': 'HUNSLET', 'companycategory': 'Private Limited Company', 'companyname': '! LTD', 'countryoforigin': None, 'regaddress_pobox': None, 'regaddress_country': None, 'dissolutiondate': None, 'regaddress_postcode': 'LS10 2RU', 'regaddress_county': 'YORKSHIRE'}], 'companynumber': ['08209948']})

我不確定這是否相關，但我注意到局部變量一旦從 dict 中取出，就會全部放在一個列表中，如下所示： ['None']但是導致問題的日期變量（ dissolutiondate ) 作為真正的None值給出。

Answer 1

好的。 所以問題原來是 psycopg2 和 postgresql 在處理數組時交互的方式，pscyopg 中曾經有一個錯誤，不允許將空值數組導入 postgres，詳細如下：

https://github.com/psycopg/psycopg2/issues/285

正如 Vao Tsun 指出的那樣，解決方案是在每個 unnest 語句的強制轉換中，這必須是顯式的，但還必須在每個數據類型說明符之后包含[]括號。

我也在這里錯誤地將我的變量轉換為 python 中的字符串：

companyname=[str(r['companyname']) for r in data]

這導致None值變成了'None'值的字符串。

這是正確代碼的示例：

SELECT
          unnest( %(companyname)s::TEXT[] ),
          unnest( %(companynumber)s::TEXT[] ),
          unnest( %(regaddress_careof)s::TEXT[] ),
          unnest( %(regaddress_pobox)s::TEXT[] ),
          unnest( %(regaddress_addressline1)s::TEXT[] ),
          unnest( %(regaddress_addressline2)s::TEXT[] ),
          unnest( %(regaddress_posttown)s::TEXT[] ),
          unnest( %(regaddress_county)s::TEXT[] ),
          unnest( %(regaddress_country)s::TEXT[] ),
          unnest( %(regaddress_postcode)s::TEXT[] ),
          unnest( %(companycategory)s::TEXT[] ),
          unnest( %(companystatus)s::TEXT[] ),
          unnest( %(countryoforigin)s::TEXT[] ),
          unnest( %(dissolutiondate)s::TIMESTAMP[] ),

和

companyname=[(r['companyname']) for r in data]
    companynumber=[(r['companynumber']) for r in data]
    regaddress_careof=[(r['regaddress_careof']) for r in data]
    regaddress_pobox=[(r['regaddress_pobox']) for r in data]
    regaddress_addressline1=[(r['regaddress_addressline1']) for r in data]
    regaddress_addressline2=[(r['regaddress_addressline2']) for r in data]
    regaddress_posttown=[(r['regaddress_posttown']) for r in data]
    regaddress_county=[(r['regaddress_county']) for r in data]
    regaddress_country=[(r['regaddress_country']) for r in data]
    regaddress_postcode=[(r['regaddress_postcode']) for r in data]
    companycategory=[(r['companycategory']) for r in data]
    companystatus=[(r['companystatus']) for r in data]
    countryoforigin=[(r['countryoforigin']) for r in data]
    dissolutiondate=[datetime.strptime(r['dissolutiondate'], "%d/%m/%Y") if r['dissolutiondate'] else None for r in data]