Terrazin

Idea

Simple and comfortable, as possible, data structures parser in to SQL.

Data

Describing sql with data structures like honeysql or ql in clojure.

Constructor

Construct data structures inside Constructor instance.

Result

Get result and access any returned data rails like syntax.

Realization

This is my first gem and first close meeting with OOP... I would appreciate any help =) And sorry for my English =(

Readiness

Terrazine is not finished yet. Now it has allmost full SELECT builder, but with some limitations like:

• awfull where syntax
• bad join syntax
• not all SQL functions supported
  

And now it supports only Postgresql.

Detailed description

Usage

Initialization

Add this line to the Gemfile

gem 'terrazine', '0.0.2'

After server initialization set Terrazine.config. Now config accepts only :connection option. In the bright future will be added :adapter option support.
In rails you can set config with after_initialize and it will looks like:

UPD: On production, rails closing PG::Connection from after_initialize, as fast fix connection now can be Proc object which must return PG::Connection. Later i'll try to find better solution

# file config/application.rb
module Name
  class Application < Rails::Application
  # ....
    config.after_initialize do
      Terrazine.config connection: -> { ActiveRecord::Base.connection.raw_connection }
    end
  # ....
  end
end

Workflow

• Describe whole data structure, or create Constructor instance and combine parts of data by it instance methods.
• Send result to Terrazine.send_request(structure||constructor, params = {})
• Rejoice at the ::Result

Constructor

You can create Constructor instance by calling Terrazine.new_constructor. It optional accepts data structure.

constructor = Terrazine.new_constructor
constructor_2 = Terrazine.new_constructor from: :calls

Instance methods

Instance methods write or combine data inside constructor instance. Not finished methods - just rewrites structure without combination with existing data.

• [ ] with
• [x] select/distinct_select
• [ ] from
• [ ] join
• [x] where
• [x] limit
• [x] paginate
• [x] merge - just merging instance structure with argument
• [x] build_sql

Data Structures

You can take a look on more detailed examples in spec/constructor_spec.rb

Common patterns

SQL Function

Structure:

• Array
  • first element - Symbol that begins from _ - :_nullif
  • arguments
  • columns
    

[:_count, [:_nullif, :row, [:_params, 'mrgl']]] # TODO: param?
# => ['COUNT(NULLIF(row, $1))', ['mrgl']]

Detailed Functions description.

Columns

Possible structures:

• String
  • if it locted in the Hash with table alias, table alias will be added to it
  • if there is no table alias it will be returned to the builder as it is.
• Symbol - just parsed to string
• Hash
  • key - table alias||name
  • value - columns
• SQL function
• Array - holder of any possible structures
  ruby ['name', {u: ['role', 'u.phone, m.rating', :field]}] # => 'name, u.role, u.phone, m.rating, u.field'
  Detailed Select description ###### Tables Possible structures:
• String || Symbol
• SQL function
• Array
  • if there is no Array inside it will be joined structure.join ' '
  • otherwise it will be recursive mapped
    ruby ['users u', [:_values, ...], [:masters, :m]] 'users u, (VALUES...), masters m'
    ###### Conditions Not finished yet... Column can be described as :u__name => 'u.name' or :name Possible structures:
• String passes as it is
• Hash represent sql = or IN if value is Array. TODO: IS in case of nil or false
  • Symbol - column name
  • Array - only as value! will be placed in querry params ($1).
  • String - will be placed in querry params
• Array
  • first element - Symbol operator representation, by default :and
  • eq - =
  • or, and
  • in
  • not
  • like, ilike
  • reg - ~, reg_i - ~*, reg_f - !~, reg_fi - !~*
  • arguments
• Array - holder of any possible structures
  ruby [[:not, 'z = 13'], [:or, 'mrgl = 2', 'rgl = 22'], [:or, 'rgl = 12', 'zgl = lol']] # => 'NOT z = 13 AND (mrgl = 2 OR rgl = 22) AND (rgl = 12 OR zgl = lol)' [{ role: 'manager', id: [0, 1, 153] }, [:not, [:like, :u__name, 'Aeonax']]] #=> 'role = $1 AND id IN ($2) AND NOT u.name LIKE $3', ['manager', [0, 1, 153], 'Aeonax']
  ###### Sub Querry Possible structures:
• Constructor instance
• Hash with :select value
  

Select

Possible structures:

• columns
• sub querry
• SQL function
• Array with combination of possible structures.
  ``ruby # String constructor.select "name, email" # Symbol constructor.select :birthdate # Array as columns constructor.select [:phone, 'role'] # Array as SQL function constructor.select [:_nullif, :row, :value] # Hash with column alias(AS) as key and any available forselectvalue constructor.select _missed_calls_count: { select: [:_count, [:_nullif, :connected, :true]], from: [:calls, :c], where: ['c.client_id = u.id', ['direction = ?', 0]]} # Hash with table alias as key and any available forselect` values constructor.select m: [:common_rating, :work_rating, { _master_id: :id }] # You can take a look of resulted data structure. In future, perhaps, Constructor will be more complicated and it will merge hashes... constructor.structure # => { select: ['name, email', :birthdate, :phone, 'role', # [:_nullif, :row, :value], # { _missed_calls_count: { select: [:_count, [:_nullif, :connected, :true]], # from: [:calls, :c], # where: ['c.client_id = u.id', # ['direction = ?', 0]]} }] }, # { m: [:common_rating, :work_rating, { _master_id: :id }] }

constructor.build_sql

=> ['SELECT name, email, birthdate, phone, role, NULLIF(row, value), m.common_rating, m.work_rating, m.id AS master_id,

(SELECT COUNT(NULLIF(connected, TRUE))

FROM calls c

WHERE c.client_id = u.id AND direction = $1) AS missed_calls_count',

0]

##### Distinct Select
To specify distinct select you should add to your data structure `:distinct` value:
- `true`
- [columns](#columns)  
Or with `Constructor` instance methods:
- `.distinct`
  - distinct structure - optional
- `.distinct_select`
  - [select](#select) structure
  - distinct structure - optional
In constructor methods `distinct: true` passed by default  
```ruby
# as data
distinct: true, select: true
# => 'SELECT DISTINCT * '
# OR via constructor
constructor.distinct_select([:id, :name]).build_sql # => 'SELECT DISTINCT id, name'
# OR
constructor.distinct_select([:id, :name], :phone).build_sql # => 'SELECT DISTINCT ON(phone) id, name '

From

Possible structures:

• table representation
• SQL functions
• Array with combination of possible structures.
  ruby from: 'table_name table_alias' || :table_name from: [:table_name, :table_alias] # => 'FROM table_name table_alias ' from: [:_values, [1, 2], :rgl, [:zgl, :gl]] # => 'FROM (VALUES(1, 2)) AS rgl (zgl, gl)' from: [[:table_name, :table_alias], [:_values, [1, 2], :values_name, [*values_column_names]]] # => 'FROM table_name table_alias, (VALUES(1, 2)) AS values_name (v_c_1, v_c_2)'
  I do not like the from syntax, but how it can be made more convenient...?

Join

Possible structures:

• String - just passed in to JOIN #{structure}
• Array with values(same order):
  • table representation
  • Hash
  • :on - conditions
  • :option - optional - contains Symbol or String of join type... rename to type?
• Array with combination of possible structures.
  ruby join: 'users u ON u.id = m.user_id' join: ['users u ON u.id = m.user_id', 'skills s ON u.id = s.user_id'] join: [[:user, :u], { on: 'rgl = 123' }] # => 'JOIN users u ON rgl = 123' join: [[[:user, :u], { option: :full, on: [:or, 'mrgl = 2', 'rgl = 22'] }], [:master, { on: ['z = 12', 'mrgl = 12'] }]] # => 'FULL JOIN user u ON mrgl = 2 OR rgl = 22 JOIN master ON z = 12 AND mrgl = 12'
  

Order

Possible structures:

• String, Symbol just insert it in "ORDER BY #{structure} "
• SQL function
• Hash
  • key - previsious possible structures.
  • value - options representation
  • Symbol - :last || :first || :asc || :desc
  • String - '<' || '>' or smthng else that passed in to USING
  • Array - with symbols inside
• Array - any possible structures
  ruby order: 'z.amount DESC' || :name # => 'ORDER BY z.amount DESC ' || 'ORDER BY name ' order: [:name, [:_case ...], { amount: [:first, :desc] }] # => 'ORDER BY name, CASE ..., amount DESC NULLS FIRST '
  

With

with: [:alias_name, { select: true, from: :users}]
with: [[:alias_name, { select: true, from: :users}],
       [:alias_name_2, { select: {u: [:name, :email]},
                        from: :rgl}]]
# => 'WITH alias_name (SELECT * FROM users ), alias_name_2 (...) '
# OR
with name: { select: true },
     another_name: { select: :mrgl }
# => 'WITH name AS (SELECT * ), another_name AS (SELECT mrgl ) '

Union

union: [{ select: true, from: [:o_list, [:_values, [1], :al, [:master]]] },
        { select: true, from: [:co_list, [:_values, [0, :FALSE, :TRUE, 0],
                                                    :al, [:rating, :rejected,
                                                          :payment, :master]]] }]
'SELECT ... UNION SELECT ...'

SQL Functions

Params

Pass argument as params to adapter

[:_values, [:_params, 'mrgl', true, 'rgl'], :z, [:f_1, :f_2, :f_3]]
['(VALUES($1, $2, $3) AS z (f_1, f_2, f_3))', ['mrgl', true, 'rgl']]

Values

Second and third arguments are nesessary right now, but in furure i'll do them optional. Arguments:

• array of values, can be nested
• AS name
• column names
  ruby [:_values, [{u: [:name, :phone]}, :role, [:_params, 'rgl']], :z, [:n, :p, :r, :m]] # => '(VALUES(u.name, u.phone, role, $1) AS z (n, p, r, m))' ### Result representation #### ::Row Result row - allow accessing data by field name via method - row.name # => "mrgl" or get hash representation with row.to_h Contains
• values
• pg_result - ::Result instance

::Result < ::Row

Data can be accessed like from row - it use first row, or you can iterate rows.
Methods each, each_with_index, first, last, map, count, present? delegates to rows. index delegates to fields.
For data representation as Hash or Array exists method present
After initialize PG::Result cleared

Contains

• rows - Array of ::Row
• fields - Array of column/alias names of returned data
• options ##### Options
• :types - hash representing which column require additional parsing and which type
• :presenter_options

::Presenter

Used in result.present(options = {}) for data representation as Hash or Array. Options are merged with result.options[:presenter_options]
Data will be presented as Array if rows > 1 or options[:array] present.

Available options

• array - if querry returns only one row, but on client you await for array of data.
• structure - Hash with field as key and value as modifier. Modifier will rewrite field value in result. Modifier acts:
  • Proc - it will call proc with row as argument, and! then pass it to modifier_presentation again
  • ::Result - it will call modifier.present
  • any else will be returned without changes
• delete - (will be soon) - Symbol, String or Array representing keys that must be deleted from result data.

TODO:

Except this todo's there is a lot commented todo's inside project.-_-

• [x] Parse data like arrays, booleans, nil to SQL. (:_params function -_-)
• [x] Relocate functions builder in to class, finally I found how it can be done nice=))
• [ ] should I bother with extra spaces?
• [ ] logger in config
• [ ] Insert
• [ ] Update
• [ ] Delete

Tests

• [ ] Normal structure!!!!
• [ ] Constructor + Builder
• [ ] Result
• [ ] Request

Meditate

• [ ] builder structure... another possibility to split it?
• [ ] from
• [ ] join !!!
• [x] where !!!!!! Supporting rails like syntax with hash?
• [ ] supporting another databases

Updates:

0.0.3

• Expand predicates syntax
• added support of multiple rows for VALUES
• ORDER structure
• scary tests-_-

Contact

You can write me your suggestions for improving the syntax, wishes, things that you think are missing here.
My email, Ruby On Rails slack