ActiveRecord的进口和放大器;在PostgreSQL的连载专栏 [英] Activerecord-import & serial column in PostgreSQL

问题描述

我在升级的Rails 2.3.4项目到Rails 3.1.1的过程。旧版本中使用AR-扩展处理数据导入。我掏出AR-扩展和与ActiveRecord的进口，这点我明白了完全相同的接口。

I am in the process of upgrading a Rails 2.3.4 project to Rails 3.1.1. The old version used ar-extensions to handle a data import. I pulled out ar-extensions and replaced it with activerecord-import, which I understand has exactly the same interfaces.

我的code调用看起来像这样

My code calls looks like this

Student.import(columns, values)

两者的args是保持正确的数据有效阵列，但我得到一个大胖子的错误！

Both args are valid arrays holding the correct data, but I get a big fat error!

错误堆栈看起来是这样的：

The error stack looks like this:

NoMethodError (You have a nil object when you didn't expect it!
You might have expected an instance of Array.
The error occurred while evaluating nil.split):
  activerecord (3.1.1) lib/active_record/connection_adapters/postgresql_adapter.rb:828:in 'default_sequence_name'
  activerecord (3.1.1) lib/active_record/base.rb:647:in `reset_sequence_name'
  activerecord (3.1.1) lib/active_record/base.rb:643:in `sequence_name'
  activerecord-import (0.2.9) lib/activerecord-import/import.rb:203:in `import'

翻翻code却仿佛ActiveRecord的进口要求的ActiveRecord从而查找名称和Postgres的序列的下一个值。

Looking through the code it seems as though Activerecord-import calls activerecord which in turn looks for the name and next value of the Postgres sequence.

所以ActiveRecord的导入查找SEQUENCE_NAME LIB / ActiveRecord的导入/ import.rb：203

So activerecord-import looks for the sequence_name lib/activerecord-import/import.rb:203

# Force the primary key col into the insert if it's not
# on the list and we are using a sequence and stuff a nil
# value for it into each row so the sequencer will fire later
-> if !column_names.include?(primary_key) && sequence_name && connection.prefetch_primary_key?
   column_names << primary_key
   array_of_attributes.each { |a| a << nil }
end

它调用活动记录... LIB / active_record / base.rb：647：在'reset_sequence_name

It calls active record ... lib/active_record/base.rb:647:in `reset_sequence_name'

# Lazy-set the sequence name to the connection's default. This method
# is only ever called once since set_sequence_name overrides it.
def sequence_name #:nodoc:
->  reset_sequence_name
end

def reset_sequence_name #:nodoc:
  -> default = connection.default_sequence_name(table_name, primary_key)
  set_sequence_name(default)
  default
end

在 serial_sequence 返回nil和 default_sequence_name 试图拆分它。

The code errors when serial_sequence returns nil and default_sequence_name tries to split it.

lib/active_record/connection_adapters/postgresql_adapter.rb

# Returns the sequence name for a table's primary key or some other specified key.
def default_sequence_name(table_name, pk = nil) #:nodoc:
  -> serial_sequence(table_name, pk || 'id').split('.').last
rescue ActiveRecord::StatementInvalid
  "#{table_name}_#{pk || 'id'}_seq"
end

def serial_sequence(table, column)
  result = exec_query(<<-eosql, 'SCHEMA', [[nil, table], [nil, column]])
    SELECT pg_get_serial_sequence($1, $2)
  eosql
  result.rows.first.first
end

当我执行 pg_get_serial_sequence（）直接对我没有得到任何返回值的数据库：

When I execute pg_get_serial_sequence() directly against the database I get no value returned:

SELECT pg_get_serial_sequence('student', 'id')

不过，我可以看到，在数据库中有一个序列名为 student_id_seq

我使用的红宝石以下版本，轨道PG等。

I am using the following versions of Ruby, rails PG etc..

• 的Rails 3.1.1
• 的Ruby 1.9.2
• ActiveRecord的进口0.2.9
• 第0.12.2提供
• 在PSQL（9.0.5，9.1.3服务器）

我已经迁移的数据库从MySQL到PostgreSQL，我不认为这对问题有任何影响，但我认为我最好添加它的完整性。

I have migrated the database from MySQL to PostgreSQL, I don't think this has any bearing on the problem but I thought that I'd better add it for completeness.

我不明白为什么这是行不通的！

I can't work out why this isn't working!

推荐答案

• 表学生的存在。
• 在列 ID 的存在。
• 顺序 student_id_seq 的存在。
• pg_get_serial_sequence（'学生'，'ID'）仍返回 NULL 。

Summary of your description:

• The table student exists.
• The column id exists.
• The sequence student_id_seq exists.
• pg_get_serial_sequence('student', 'id') still returns NULL.

1）顺序为不挂以列。

默认列与列和序列之间的联系是独立的功能。仅仅存在一个合适的顺序并不意味着它做什么，你presume。如果您创建一个列系列你全包，但。阅读在手动的细节。

Column default and the tie between column and sequence are independent features. The mere existence of a fitting sequence does not mean it does what you presume. If you create a column as serial you get the whole package, though. Read the details in the manual.

要解决这个问题（如果你确信这是应该的），你可以标记的顺序为归 student.id ：

To fix this (and if you are sure that's how it should be), you can mark the sequence as "owned by" student.id:

ALTER SEQUENCE student_id_seq OWNED BY student.id;

另外检查列默认设置为预期的：

Also check if the column default is set as expected:

SELECT column_name, column_default
FROM   information_schema.columns
WHERE  table_name = 'student'
-- AND    schema = 'your_schema' -- if needed

如果没有，修：

ALTER TABLE student ALTER COLUMN id SET DEFAULT nextval('student.id')

2）的的mixup 表名的主机地址/端口/数据库/模式/资本化。

2) A mixup of host address / port / database / schema / capitalization of the table name.

这一切发生的时间。请确保您检查您的应用程序连接到同一个数据库，使用相同的用户或者至少是相同的 SEARCH_PATH 。确保的对象是在你所期望的模式并没有，比如，另一个学生表中的搞混了另一种模式。

It happens all the time. Make sure you check the same database that your app connects to, With the same user or at least the same search_path. Make sure, the objects are in the schema where you expect them and there isn't, for instance, another student table in another schema that got mixed up.

这篇关于ActiveRecord的进口和放大器;在PostgreSQL的连载专栏的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！