Convert PostgreSQL to SQLite

Convert PostgreSQL to SQLite

1. Dump the data only sql to file
  $ pg_dump --data-only --inserts YOUR_DB_NAME > dump.sql
2. scp to local

3. Remove the SET statements at the top
  such as:
    SET statement_timeout = 0;
    SET client_encoding = 'SQL_ASCII';

4. Remove the setval sequence queries
  such as: 
    SELECT pg_catalog.setval('MY_OBJECT_id_seq', 10, true);
    
5. Replace true => ‘t’ and false => ‘f’
  -- These:
  INSERT INTO table_name VALUES (1, true, false);
  -- Should be replace to:
  INSERT INTO table_name VALUES (1, 't', 'f');

6. Add BEGIN; and END; to wrap the whole thing as a trasaction

7. Import
  $ rm db/development.sqlite3
  $ rake db:migrate
  $ sqlite3 db/development.sqlite3
  > delete from schema_migrations;
  > .read dump.sql
  > .exit

This was very helpful but may be out of date now.
I found I had to remove "public" prefix from table names, and instead of "t"/"f" I could use true/false values per the export.

Not sure if it was just my sequence but also performed:
delete from ar_internal_metadata;

(along with delete from schema_migrations;)

sqlite> .version
SQLite 3.39.5 2022-10-14 20:58:05 554764a6e721fab307c63a4f98cd958c8428a5d9d8edfde951858d6fd02daapl
clang-14.0.3

Just to wrap this up I had to change some postgresql JSONB fields to JSON, and some Integer/String Array fields to JSON fields.

I also had to use different date calculation queries using sqlite's JULIANDAY and the TRIM functions were different in my search/filter queries and indexes. But that was pretty much it!

I turned this into a simple, but effective bash script:
https://github.com/hirefrank/pg-to-sqlite

I'm on Postgresql 16 and the default dump file has non compatible format of INSERT INTO with the one sqlite uses. I found out if you dump DB with --attribute-inserts it will work

pg_dump --attribute-inserts --data-only -d my_psql_db > /tmp.dump.sql

reference pg_dump docs

https://hhh.how/notes/2067-convert-postgresql-to-sqlite

This was super helpful. Thank you!

I'm migrating a Rails database, and thought it worth adding that the INSERTS for both the ar_internal_metadata table and schema_migrations table should be removed as these are created when running rails db:migrate.

They will raise UNIQUE constraint failed errors otherwise.

Thanks @equivalent and @aaronmoodie I added your suggestions to my script -- you can see them here:
hirefrank/pg-to-sqlite#1

Also @aaronmoodie 👋 it's been a long time since Etsy!

Oh! hey @hirefrank! 👋 I didn't realise that was you! Thanks for updating the script and for adding my suggestions.

@hirefrank you forgot to remove pg_catalog occurrences

Thanks @jvitoroc! Pls review and rest: hirefrank/pg-to-sqlite#3

If this works, I'll merge!

Merged!

such a clusterfuck....

it's not straight forward, if you use enums, lots of timestamps etc.
the db dump is just not compatible.

solution: dump to raw data, and load from raw data.

add this to your ApplicationRecord or modify whatever you want

  def self.to_csv
    file_path = Rails.root.join('db', 'seeds', "#{self.name.underscore.pluralize}.csv")

    FileUtils.mkdir_p(File.dirname(file_path)) unless File.directory?(File.dirname(file_path))

    CSV.open(file_path, 'wb') do |csv|
      csv << column_names
      self.unscoped.each do |record|
        csv << record.attributes.values
      end
    end
    true
  end

  def self.from_csv
    file_path = Rails.root.join('db', 'seeds', "#{self.name.underscore.pluralize}.csv")

    raise "file no exista" unless File.exist?(file_path)
    rows = []
    CSV.foreach(file_path, headers: true) do |row|
      attributes = row.to_hash

      column_types = self.columns_hash

      column_types.each do |column_name, column_info|
        if column_info.type == :datetime || column_info.type == :timestamp
          if attributes[column_name]
            attributes[column_name] = Time.parse(attributes[column_name])
          end
        end
      end

      rows << attributes
    end

    if rows.any?
      self.insert_all(rows)
      puts "#{rows.size} records imported successfully!"
    end
  end


   def self.dump_all_models_to_csv
    descendants.each do |model|
      next unless model.descendants.empty?
      model.to_csv
      puts "Exported #{model.name} to CSV successfully."
    end
  end


  def self.import_all_models_from_csv
    seeds_path = Rails.root.join('db', 'seeds')
    csv_files = Dir.glob(seeds_path.join('*.csv'))

    csv_files.each do |file_path|
      file_name = File.basename(file_path, '.csv')
      model_name = file_name.singularize.camelize
      if Object.const_defined?(model_name)
        model = model_name.constantize
        model.unscoped.delete_all
        model.from_csv
        puts "Imported data from #{file_name}.csv successfully."
      else
        puts "No model found for #{file_name}.csv; skipping import."
      end
    end
  end

Specifically for use with https://github.com/redwoodjs/sdk (which uses Cloudflare D1) when migrating from https://github.com/redwoodjs/graphql/ (and using Postgres), I created the following script based on @hirefrank 's script:

https://gist.github.com/arimendelow/e60f8a1303dc00bd4a21e84142623dbc

hirefrank/pg-to-sqlite#5 lmk what you think @arimendelow

oooo will take a look @hirefrank !!! i'll reply back on your PR :)