Skip to content

Latest commit

 

History

History
269 lines (190 loc) · 8.85 KB

README.md

File metadata and controls

269 lines (190 loc) · 8.85 KB
Raw

postgres-copy

Ruby

This Gem will enable your AR models to use the PostgreSQL COPY command to import/export data in CSV format. If you need to tranfer data between a PostgreSQL database and CSV files, the PostgreSQL native CSV parser will give you a greater performance than using the ruby CSV+INSERT commands. I have not found time to make accurate benchmarks, but in the use scenario where I have developed the gem I have had a four-fold performance gain. This gem was written having the Rails framework in mind, I think it could work only with active-record, but I will assume in this README that you are using Rails.

Install

Put it in your Gemfile

gem 'postgres-copy'

Run the bundle command

bundle

IMPORTANT note about recent versions

  • Rails 4 users should use the version 0.7 and onward, while if you use Rails 3.2 stick with the 0.6 versions.
  • Since version 0.8 all methods lost the prefix pg_ and they should be included in models thourgh acts_as_copy_target.

Usage

To enable the copy commands in an ActiveRecord model called User you should use:

class User < ActiveRecord::Base
  acts_as_copy_target
end

This will add the additional class methods to your model:

  • copy_to
  • copy_to_string
  • copy_to_enumerator
  • copy_from

Using copy_to and copy_to_string

You can go to the rails console and try some cool things first. The first and most basic use case, let's copy the entire content of a database table to a CSV file on the database server disk. Assuming we have a users table and a User AR model:

User.copy_to '/tmp/users.csv'

This will execute in the database the command:

COPY (SELECT "users".* FROM "users" ) TO '/tmp/users.csv' WITH DELIMITER ',' CSV HEADER

Remark that the file will be created in the database server disk. But what if you want to write the lines in a file on the server that is running Rails, instead of the database? In this case you can pass a block and retrieve the generated lines and then write them to a file:

File.open('/tmp/users.csv', 'w') do |f|
  User.copy_to do |line|
    f.write line
  end
end

Instead of yielding each line, you could return an enumerator with all users:

enumerator = User.copy_to_enumerator

And for better performance when rendering the result of the enumerator, you can return an enumerator with blocks of 100 lines joined:

enumerator = User.copy_to_enumerator(:buffer_lines => 100)

Or, if you have enough memory, you can read all table contents to a string using .copy_to_string

puts User.copy_to_string

Another interesting feature of copy_to is that it uses the scoped relation, it means that you can use ARel operations to generate different CSV files according to your needs. Assuming we want to generate a file only with the names of users 1, 2 and 3:

User.select("name").where(:id => [1,2,3]).copy_to "/tmp/users.csv"

Which will generate the following SQL command:

COPY (SELECT name FROM "users" WHERE "users"."id" IN (1, 2, 3)) TO '/tmp/users.csv' WITH DELIMITER ',' CSV HEADER

Alternatively, you can supply customized raw SQL query to copy_to instead of scoped relation:

User.copy_to("/tmp/users.csv", query: 'SELECT count(*) as Total FROM users')

Which will generate the following SQL command:

COPY (SELECT count(*) as Total FROM users) TO '/tmp/users.csv' WITH DELIMITER ',' CSV HEADER

The COPY command also supports exporting the data in binary format.

User.select("name").where(:id => [1,2,3]).copy_to "/tmp/users.dat", :format => :binary

Which will generate the following SQL command:

COPY (SELECT name FROM "users" WHERE "users"."id" IN (1, 2, 3)) TO '/tmp/users.dat' WITH BINARY

The copy_to_string method also supports this

puts User.copy_to_string(:format => :binary)

Using copy_from

Now, if you want to copy data from a CSV file into the database, you can use the copy_from method. It will allow you to copy data from an arbritary IO object or from a file in the database server (when you pass the path as string). Let's first copy from a file in the database server, assuming again that we have a users table and that we are in the Rails console:

User.copy_from "/tmp/users.csv"

This command will use the headers in the CSV file as fields of the target table, so beware to always have a header in the files you want to import. If the column names in the CSV header do not match the field names of the target table, you can pass a map in the options parameter.

User.copy_from "/tmp/users.csv", :map => {'name' => 'first_name'}

In the above example the header name in the CSV file will be mapped to the field called first_name in the users table. You can also manipulate and modify the values of the file being imported before they enter into the database using a block:

User.copy_from "/tmp/users.csv" do |row|
  row[0] = "fixed string"
end

The above example will always change the value of the first column to "fixed string" before storing it into the database. For each iteration of the block row receives an array with the same order as the columns in the CSV file.

To specify NULL value you can pass the null option parameter.

User.copy_from "/tmp/users.csv", :null => 'null'

Match the specified columns' values against the null string, even if it has been quoted, and if a match is found set the value to NULL (Postgres 9.4+ only).

User.copy_from "/tmp/users.csv", :null => '', :force_null => [:name, :city]

To copy from tsv file , you can set format :tsv

User.copy_from "/tmp/users.tsv", :format => :tsv

To copy a binary formatted data file or IO object you can specify the format as binary

User.copy_from "/tmp/users.dat", :format => :binary

NOTE: Columns must line up with the table unless you specify how they map to table columns.

To specify how the columns will map to the table you can specify the :columns option

User.copy_from "/tmp/users.dat", :format => :binary, :columns => [:id, :name]

Which will generate the following SQL command:

COPY users (id, name) FROM '/tmp/users.dat' WITH BINARY

To specify the encoding with which to read the file, set the :encoding option. This is useful for removing byte order marks when matching column headers.

User.copy_from "/tmp/users_with_byte_order_mark.csv", :encoding => 'bom|utf-8'

Using PostgresCopy::WithTempTable.generate

Based on nfedyashev's comment:

    PostgresCopy::WithTempTable.generate do |t|
      columns.each do |column_name|
        t.string column_name.to_sym
      end
    end

This auto-generates an id column, but the temp table creation is configurable.

Using the CSV Responder

If you want to make the result of a COPY command available to download this gem provides a CSV responder that, in conjunction with inherited_resources, is a very powerfull tool. BTW, do not try to use the responder without inherited_resources.

The advantage of using this CSV responder over generating the CSV file, is that it will iterate over the resulting lines and send them to the client as they became available, so you will not have any problem with memory consumption or timeouts executing the action.

Here's an example of controller using inherited resources and the CSV Responder:

class Clients < ApplicationController
  inherit_resources
  responders :csv
  respond_to :csv
  actions :index
end

The example above should be enough to generate a CSV file with clients (given that database and routes are set up accordingly) if you access the index action passing csv as the format. The has_scope gem (included by default in inherited_resources) is very useful to filter the generated CSV file using model scopes.

You could use something such as:

class Clients < ApplicationController
  inherit_resources
  responders :csv
  respond_to :csv
  actions :index
  has_scope :by_name
end

To filter clients by name using a corresponding model scope. One could argue that this responder should be in a separate gem. But it's a very small piece of code and make the CSV available for download is a very common use case of postgres-copy.

Note on Patches/Pull Requests

  • Fork the project.
  • Make your feature addition or bug fix.
  • Add tests for it. This is important so I don't break it in a future version unintentionally.
  • Commit, do not mess with rakefile, version, or history. (if you want to have your own version, that is fine but bump version in a commit by itself I can ignore when I pull)
  • Send me a pull request. Bonus points for topic branches.

Copyright

Copyright (c) 2011 Diogo Biazus. See LICENSE for details.