datasiphon 0.2.16

Dynamic building of filtered database queries

Datasiphon

Package for applying dictionary filter to some form of query on database to retrieve filtered data or acquire filtered query

Installation

Use the package manager pip to install datasiphon.

pip install datasiphon

Usage

from siphon import sql
import sqlalchemy as sa
# Create a filter
filter_ = {
    "name": {"eq": "John"},
}

table = sa.Table("users", sa.MetaData(), autoload=True, autoload_with=engine)
# Build a query
query = table.select()
# apply filter using build function
query = sql.SQL.build(query, filter_)
# execute query
result = engine.execute(query)
...

Supported Database types

SQL package (No ORM)

• implemented using sqlalchemy package, expected to work with Table and Select objects

Building query

1. Prerequisite
   • base SELECT query (Select object) from actual Table objects (not text objects)
   • filter (dictionary), optional, optimally parsed using qstion package -> similiar to npm's qs package
   • restriction model (child of siphon.sql.RestrictionModel class), optional, to restrict the filter to certain fields
2. Usage

from siphon import sql

# Create a filter with strict form
filter_ = {
    "name": {"eq": "John"},
}

# build a query with filter
new_query = sql.SQL.build(query, filter_)

• filter_ is validated before building the query, against columns used in select statement and restriction model (if provided)
• allowed format represents nestings containing one of :

1. junctions (AND, OR) -> for combining multiple conditions with desired logical operators (allowed exclusively per nest level)

       # Example correct - joining or with different fields
       filter_ = {
           "or":
           {
               "name": {"eq": "John"},
               "age": {"gt": 20}
           }
       }
       
       # example correct - joining or with same field, different operators
       filter_ = {
           "name": {
               "or": {
                   "eq": "John",
                   "ne": "John"
               }
           }
       }
       # Example - incorrect - multiple junctions in same nest level
       filter_ = {
           "or":
           {
               "name": {"eq": "John"},
               "age": {"gt": 20}
           },
           "and":
           {
               "name": {"eq": "John"},
               "age": {"gt": 20}
           }
       }
   

2. operators (eq, ne...) -> for applying conditions on fields -> must always follow a field name (not directly but always has to be nested deeper than field name)

   # Example correct - applying eq operator on field name
   filter_ = {
       "name": {"eq": "John"}
       }
   
   # Example - incorrect - applying eq operator before field name
   filter_ = {
       "eq": {
           "name": "John"
       }
   }
   

3. field name -> for applying conditions on fields -> must always contain an operator (not directly but always has to be nested deeper than field name)

   # Example correct - applying eq operator on field name
   filter_ = {
       "name": {"eq": "John"}
       }
   
   # Example - incorrect - applying eq operator before field name
   filter_ = {
       "eq": {
           "name": "John"
       }
   }
   

• if using restriction model, filter also can contain only fields in the restriction model, and operators used have to be in restriction model's field (RestrictionModel class has pretty strict validation for initialization and annotation and won't allow any faulty initialization or usage)

  # Example of correct restriction model usage
  class UserRestrictionModel(sql.RestrictionModel):
      # must always be annotated as `list[str]` otherwise will raise an error
      name = list[str] = []
      # must be always subset of all existing operators for that query builder
      age = list[str] = ['eq']
  
  # Example incorrect - using operator not in restriction model
  class UserRestrictionModel(sql.RestrictionModel):
      # does not have correct annotation
      name = list[int] = []
      # does not have correct operators
      age = list[str] = ['eql']
  

• using multiple condition without specifying junctions will result in an AND junction between them

  # Example correct - applying eq operator on field name
  filter_ = {
      "name": {"eq": "John"},
      "age": {"gt": 20}
      }
  # will be treated as
  filter_ = {
      "and": {
          "name": {"eq": "John"},
          "age": {"gt": 20}
      }
  }
  
  filter_ = {
      "name": {
          "eq": "John",
          "ne": "John"
          }
  }
  # will be treated as
  filter_ = {
      "and": {
          "name": {
              "eq": "John",
              "ne": "John"
          }
      }
  }
  

• generating query: recursively collecting items from filter, and applying filtering directly to exported columns of given query

Reconstructing dict filter from query

• Since data that are selected using filtered queries builded by this package allow very specific filtration and juctions. Usually it is hard to reconstruct the filter from the query for possible further use in case there are more data than expected and we want to paginate them.
• For this purpose, a separate class PaginationBuilder, that serve as a helper for some trivial methods for getting specific parts/structures of the query is provided.
• Following useful methods are implemented:

1. is_query_paginable - query is paginable if the column by which is query originally ordered is not present in whereclause with direct comparison operator (eq, ne)
2. reconstruct_filter - allows you to reconstruct (functionally) equivalent filter from the query (core SQL) back into dictionary filter in qstion format
3. get_referenced_column - allows you to retrieve Column or ReadOnlyColumn from more complex queries either by its direct name or reference (label) in the query
4. retrieve_order_by - allows you to retrieve the column name(s) and direction by which the query is ordered in form of pairs (direction, column_name)
5. retrieve_filtered_column - allows you to retrieve column which is filtered in the query (if present) in form of Operation object