🤏🏻 Criteria Pattern

The Criteria Pattern is a Python 🐍 package that simplifies and standardizes criteria based filtering 🤏🏻, validation and selection. This package provides a set of prebuilt 👷🏻 objects and utilities that you can drop into your existing projects and not have to implement yourself.

These utilities 🛠️ are useful when you need complex filtering logic. It also enforces 👮🏻 best practices so all your filtering processes follow a uniform standard.

Easy to install and integrate, this is a must have for any Python developer looking to simplify their workflow, enforce design patterns and use the full power of modern ORMs and SQL 🗄️ in their projects 🚀.

Table of Contents

• 📥 Installation
• 📚 Documentation
• ✨ Features
• 💻 Utilization
  • 🧱 Core Concepts
  • 🧮 Supported Operators
  • 🔄 Available Converters
  • 🔐 Security For User-Facing APIs
  • 🛡️ SQL Conversion And Safety
  • 📦 Request Body Examples
  • 🧭 Structured URL Query Examples
  • 🎯 Real-Life Case: Multi-tenant User Search Service
  • 🔎 Simple URL Query Examples
  • 🧪 Testing Helpers
• 🤝 Contributing
• 🔑 License

🔼 Back to top

📥 Installation

You can install Criteria Pattern using pip:

pip install criteria-pattern

You can install the companion AI-agent skill from skills.sh with Vercel's skills CLI:

npx skills add adriamontoto/criteria-pattern

Review the skill source in skills/criteria-pattern before installing it in sensitive environments.

🔼 Back to top

📚 Documentation

The root README is the entry point. Deeper guides live in this repository and are linked here:

• 🧱 Usage Guide: Core models, composition rules, operators and pagination.
• 🔄 Converter Guide: SQL converters, request converters, placeholder styles and mapping.
• 🔐 Security Guide: User-facing criteria safety and injection prevention.
• 🧪 Testing Guide: Object mother helpers and testing recommendations.
• AI Skill: Installable skill package that teaches AI agents how to use Criteria Pattern safely.

This project's DeepWiki documentation is also available for generated repository navigation.

🔼 Back to top

✨ Features

Criteria Pattern gives you a typed, reusable way to describe queries before deciding how they should be executed.

• 🧱 Composable criteria objects with filters, orders, page size and page number.
• 🔗 Boolean composition with &, | and ~ for AND, OR and NOT logic.
• 🔎 20 filter operators covering equality, comparison, pattern matching, ranges, null checks and list membership.
• ↕️ Ordering primitives with duplicate order-field protection.
• 📄 Pagination primitives that enforce positive integer values and consistent page-number usage.
• 🗄️ SQL converters for PostgreSQL, MySQL, MariaDB and SQLite.
• 🧾 Request converters for decoded request bodies, structured URL queries and compact suffix-based URL queries.
• 🛡️ Security-oriented allowlist validation for tables, columns, fields, operators, directions and pagination bounds.
• 🧰 Field and operator mapping so public API names can be translated to internal model or database names.
• 🧪 Object mother testing helpers for downstream projects that want realistic criteria fixtures.

🔼 Back to top

💻 Utilization

from criteria_pattern import Criteria, Filter, Operator
from criteria_pattern.converters import CriteriaToPostgresqlConverter

is_adult = Criteria(filters=[Filter(field='age', operator=Operator.GREATER_OR_EQUAL, value=18)])
email_is_gmail = Criteria(filters=[Filter(field='email', operator=Operator.ENDS_WITH, value='@gmail.com')])
email_is_yahoo = Criteria(filters=[Filter(field='email', operator=Operator.ENDS_WITH, value='@yahoo.com')])

query, parameters = CriteriaToPostgresqlConverter.convert(
    criteria=is_adult & (email_is_gmail | email_is_yahoo),
    table='user',
    valid_columns=['age', 'email'],
    valid_operators=[Operator.GREATER_OR_EQUAL, Operator.ENDS_WITH],
)
print(query)
print(parameters)
# >>> SELECT * FROM "user" WHERE ("age" >= %(parameter_0)s AND ("email" LIKE '%%' || %(parameter_1)s OR "email" LIKE '%%' || %(parameter_2)s));
# >>> {'parameter_0': 18, 'parameter_1': '@gmail.com', 'parameter_2': '@yahoo.com'}

🧱 Core Concepts

The main model is Criteria. A criteria can contain filters, orders and pagination:

from criteria_pattern import Criteria, Direction, Filter, Operator, Order


criteria = Criteria(
    filters=[
        Filter(field='status', operator=Operator.EQUAL, value='ACTIVE'),
        Filter(field='age', operator=Operator.GREATER_OR_EQUAL, value=18),
    ],
    orders=[
        Order(field='created_at', direction=Direction.DESC),
    ],
    page_size=20,
    page_number=1,
)

Criteria can be composed without losing the boolean structure:

from criteria_pattern import Criteria, Filter, Operator


is_active = Criteria(filters=[Filter(field='status', operator=Operator.EQUAL, value='ACTIVE')])
is_adult = Criteria(filters=[Filter(field='age', operator=Operator.GREATER_OR_EQUAL, value=18)])
has_company_email = Criteria(filters=[Filter(field='email', operator=Operator.ENDS_WITH, value='@acme.com')])

criteria = is_active & (is_adult | has_company_email)
not_archived = ~Criteria(filters=[Filter(field='archived_at', operator=Operator.IS_NOT_NULL, value=None)])

Pagination is optional. page_size can be used alone for LIMIT, while page_number must be used with page_size so converters can calculate the offset.

🔼 Back to top

🧮 Supported Operators

🔼 Back to top

🔄 Available Converters

The package includes converters for SQL generation and request parsing:

• criteria_pattern.converters.CriteriaToPostgresqlConverter: Converts a Criteria object into PostgreSQL SQL + parameters.
• criteria_pattern.converters.CriteriaToMysqlConverter: Converts a Criteria object into MySQL SQL + parameters.
• criteria_pattern.converters.CriteriaToMariadbConverter: Converts a Criteria object into MariaDB SQL + parameters.
• criteria_pattern.converters.CriteriaToSqliteConverter: Converts a Criteria object into SQLite SQL + parameters.
• criteria_pattern.converters.SimpleUrlToCriteriaConverter: Parses simple public URL query parameters into a Criteria object.
• criteria_pattern.converters.UrlToCriteriaConverter: Parses URL query parameters into a Criteria object.
• criteria_pattern.converters.BodyToCriteriaConverter: Parses decoded request bodies into a Criteria object.

SQL converter output uses the placeholder style expected by each database family:

🔼 Back to top

🔐 Security For User-Facing APIs

When criteria comes from a URL, JSON body, form, dashboard or any other user-facing surface, treat every field, operator, direction and pagination value as untrusted.

Criteria Pattern parameterizes filter values for SQL converters, quotes SQL identifiers per dialect, and enables allowlist validation by default. SQL identifiers still cannot be safely parameterized by database drivers, so for user-facing APIs you must pass explicit valid_* allowlists instead of relying on implicit defaults derived from the current criteria.

Use this rule of thumb:

The safest user-facing flow is:

1. Keep allowlists in application code, not in request data.
2. Parse request input with BodyToCriteriaConverter, UrlToCriteriaConverter or SimpleUrlToCriteriaConverter.
3. Map public field names to internal field or column names with fields_mapping.
4. Enable field, operator, direction and pagination validation in the request converter.
5. Enable table, column, criteria, operator, direction and pagination validation again in the SQL converter.

from criteria_pattern import Direction, Operator
from criteria_pattern.converters import BodyToCriteriaConverter, CriteriaToPostgresqlConverter


body = {
    'filters': [{'field': 'q', 'operator': 'contains', 'value': 'Doe'}],
    'orders': [{'field': 'created', 'direction': 'DESC'}],
    'page_size': 20,
    'page_number': 1,
}

fields_mapping = {'q': 'name', 'created': 'created_at'}
valid_fields = ['name', 'created_at']
valid_operators = [Operator.CONTAINS]
valid_directions = [Direction.DESC]

criteria = BodyToCriteriaConverter.convert(
    body=body,
    fields_mapping=fields_mapping,
    check_field_injection=True,
    check_operator_injection=True,
    check_direction_injection=True,
    check_pagination_bounds=True,
    valid_fields=valid_fields,
    valid_operators=valid_operators,
    valid_directions=valid_directions,
    max_page_size=100,
    max_page_number=1000,
)

query, parameters = CriteriaToPostgresqlConverter.convert(
    criteria=criteria,
    table='users',
    columns=['id', 'name', 'created_at'],
    check_table_injection=True,
    check_column_injection=True,
    check_criteria_injection=True,
    check_operator_injection=True,
    check_direction_injection=True,
    check_pagination_bounds=True,
    valid_tables=['users'],
    valid_columns=['id', 'name', 'created_at'],
    valid_operators=valid_operators,
    valid_directions=valid_directions,
    max_page_size=100,
    max_page_number=1000,
)

This keeps the public API flexible while ensuring SQL identifiers and query behavior are constrained by code you control.

🔼 Back to top

🛡️ SQL Conversion And Safety

Filter values are parameterized, which protects the values passed to predicates. Identifier validation is separate and must be enabled explicitly when identifiers can be influenced by users.

from criteria_pattern import Criteria, Direction, Filter, Operator, Order
from criteria_pattern.converters import CriteriaToPostgresqlConverter


criteria = Criteria(
    filters=[Filter(field='public_name', operator=Operator.CONTAINS, value='Doe')],
    orders=[Order(field='created_at', direction=Direction.DESC)],
    page_size=20,
    page_number=2,
)

query, parameters = CriteriaToPostgresqlConverter.convert(
    criteria=criteria,
    table='users',
    columns=['id', 'name', 'email'],
    columns_mapping={'public_name': 'name'},
    check_table_injection=True,
    check_column_injection=True,
    check_criteria_injection=True,
    check_operator_injection=True,
    check_direction_injection=True,
    check_pagination_bounds=True,
    valid_tables=['users'],
    valid_columns=['id', 'name', 'email', 'created_at', 'public_name'],
    valid_operators=[Operator.CONTAINS],
    valid_directions=[Direction.DESC],
    max_page_size=100,
    max_page_number=1000,
)

print(query)
print(parameters)
# >>> SELECT "id", "name", "email" FROM "users" WHERE "name" LIKE '%%' || %(parameter_0)s || '%%' ORDER BY "created_at" DESC LIMIT %(limit_1)s OFFSET %(offset_2)s;
# >>> {'parameter_0': 'Doe', 'limit_1': 20, 'offset_2': 20}

Use columns_mapping when your public API fields should not expose your database column names directly. When check_criteria_injection is enabled, include the accepted public criteria field names in valid_columns too.

🔼 Back to top

📦 Request Body Examples

Use BodyToCriteriaConverter when your API receives a decoded dictionary, for example from JSON.

from criteria_pattern import Direction, Operator
from criteria_pattern.converters import BodyToCriteriaConverter


body = {
    'filters': [
        {'field': 'full_name', 'operator': 'contains', 'value': 'Doe'},
        {'field': 'status', 'operator': 'IN', 'value': ['ACTIVE', 'PENDING']},
        {'field': 'price', 'operator': 'BETWEEN', 'value': [10, 100]},
    ],
    'orders': [
        {'field': 'created_at', 'direction': 'desc'},
    ],
    'page_size': 20,
    'page_number': 1,
}

criteria = BodyToCriteriaConverter.convert(
    body=body,
    fields_mapping={'full_name': 'name'},
    operator_mapping={'after': Operator.GREATER},
    check_field_injection=True,
    check_operator_injection=True,
    check_direction_injection=True,
    valid_fields=['name', 'status', 'price', 'created_at'],
    valid_operators=[Operator.CONTAINS, Operator.IN, Operator.BETWEEN],
    valid_directions=[Direction.DESC],
)

print(criteria.filters[0].field, criteria.filters[0].operator, criteria.filters[0].value)
# >>> name CONTAINS Doe

Accepted body keys are filters, orders, page_size and page_number. Unknown keys, missing required filter/order keys and invalid value shapes raise IntegrityError.

🔼 Back to top

🧭 Structured URL Query Examples

Use UrlToCriteriaConverter when you want an explicit URL format that can express filters, orders and pagination.

from criteria_pattern import Direction, Operator
from criteria_pattern.converters import UrlToCriteriaConverter


url = (
    'https://api.example.com/users?'
    'filters[0][field]=name&filters[0][operator]=CONTAINS&filters[0][value]=Doe&'
    'filters[1][field]=age&filters[1][operator]=GREATER_OR_EQUAL&filters[1][value]=18&'
    'orders[0][field]=created_at&orders[0][direction]=DESC&'
    'page_size=20&page_number=1'
)

criteria = UrlToCriteriaConverter.convert(
    url=url,
    valid_fields=['name', 'age', 'created_at'],
    valid_operators=[Operator.CONTAINS, Operator.GREATER_OR_EQUAL],
    valid_directions=[Direction.DESC],
)

print(criteria.filters[0].field, criteria.filters[0].operator, criteria.filters[0].value)
# >>> name CONTAINS Doe

print(criteria.orders[0].field, criteria.orders[0].direction)
# >>> created_at DESC

Structured URL values are converted to useful primitive types where possible: booleans, null / none, integers and floats.

🔼 Back to top

🎯 Real-Life Case: Multi-tenant User Search Service

Imagine an admin dashboard where each request must:

1. Always restrict results to the current tenant.
2. Optionally filter active users.
3. Search only users with company emails.
4. Sort by newest users first.

With Criteria Pattern, each concern is a small reusable criteria object. You combine them using & and |, then convert once to SQL:

from criteria_pattern import Criteria, Direction, Filter, Operator, Order
from criteria_pattern.converters import CriteriaToPostgresqlConverter


class UserSearchService:
    def __init__(self, tenant_id: str) -> None:
        self.tenant_id = tenant_id

    def build_query(self, *, only_active: bool, corporate_domain: str) -> tuple[str, dict[str, object]]:
        tenant_scope = Criteria(filters=[Filter(field='tenant_id', operator=Operator.EQUAL, value=self.tenant_id)])
        active_scope = Criteria(filters=[Filter(field='is_active', operator=Operator.EQUAL, value=True)])
        email_scope = Criteria(filters=[Filter(field='email', operator=Operator.ENDS_WITH, value=corporate_domain)])
        sort_scope = Criteria(orders=[Order(field='created_at', direction=Direction.DESC)])

        criteria = tenant_scope & email_scope & sort_scope
        if only_active:
            criteria = criteria & active_scope

        return CriteriaToPostgresqlConverter.convert(
            criteria=criteria,
            table='users',
            valid_columns=['tenant_id', 'is_active', 'email', 'created_at'],
            valid_operators=[Operator.EQUAL, Operator.ENDS_WITH],
            valid_directions=[Direction.DESC],
        )


service = UserSearchService(tenant_id='tenant_123')
query, parameters = service.build_query(only_active=True, corporate_domain='@acme.com')

print(query)
print(parameters)

🔼 Back to top

🔎 Simple URL Query Examples

Use SimpleUrlToCriteriaConverter when you want a compact public query format where each parameter becomes one AND filter. Plain parameters use equality, and suffixes map to operators.

from criteria_pattern import Operator
from criteria_pattern.converters import SimpleUrlToCriteriaConverter


criteria = SimpleUrlToCriteriaConverter.convert(
    url='https://api.example.com/users?name=Doe&age_gte=18&page_size=20&page_number=1',
    valid_fields=['name', 'age'],
    valid_operators=[Operator.EQUAL, Operator.GREATER_OR_EQUAL],
)

print(criteria.filters[0].field, criteria.filters[0].operator, criteria.filters[0].value)
# >>> name EQUAL Doe

print(criteria.filters[1].field, criteria.filters[1].operator, criteria.filters[1].value)
# >>> age GREATER_OR_EQUAL 18

print(criteria.page_size, criteria.page_number)
# >>> 20 1

Common suffixes:

Comma-separated values are also supported for list operators:

criteria = SimpleUrlToCriteriaConverter.convert(
    url='https://api.example.com/users?status_in=ACTIVE,PENDING,BLOCKED',
    valid_fields=['status'],
    valid_operators=[Operator.IN],
)

print(criteria.filters[0].value)
# >>> ['ACTIVE', 'PENDING', 'BLOCKED']

You can map public field names to internal field names:

criteria = SimpleUrlToCriteriaConverter.convert(
    url='https://api.example.com/users?full_name_contains=Doe',
    fields_mapping={'full_name': 'name'},
    valid_fields=['name'],
    valid_operators=[Operator.CONTAINS],
)

print(criteria.filters[0].field, criteria.filters[0].operator, criteria.filters[0].value)
# >>> name CONTAINS Doe

You can also extend or override URL suffixes:

criteria = SimpleUrlToCriteriaConverter.convert(
    url='https://api.example.com/users?created_at_after=2026-05-18',
    suffix_operator_mapping={'after': Operator.GREATER},
    valid_fields=['created_at'],
    valid_operators=[Operator.GREATER],
)

print(criteria.filters[0].field, criteria.filters[0].operator, criteria.filters[0].value)
# >>> created_at GREATER 2026-05-18

🔼 Back to top

🧪 Testing Helpers

The package includes object mother helpers for tests in downstream projects. They are useful when you need valid random criteria objects and want to override only the fields that matter for a specific test.

from criteria_pattern import Filter, Operator
from criteria_pattern.models.testing.mothers import CriteriaMother
from criteria_pattern.models.testing.mothers.filter import FilterMother


criteria = CriteriaMother.with_filters(
    filters=[
        Filter(field='status', operator=Operator.EQUAL, value='ACTIVE'),
        FilterMother.create(field='age', operator=Operator.GREATER_OR_EQUAL, value=18),
    ]
)

print(criteria.has_filters())
# >>> True

Available helpers include CriteriaMother, FilterMother, FiltersMother, OrderMother, OrdersMother, PageSizeMother, PageNumberMother, plus field/operator/direction mothers under the filter and order testing packages.

🔼 Back to top

🤝 Contributing

We love community help! Before you open an issue or pull request, please read:

• 🤝 How to Contribute
• 🧭 Code of Conduct
• 🔐 Security Policy

Thank you for helping make 🤏🏻 Criteria Pattern package awesome! 🌟

🔼 Back to top

🔑 License

This project is licensed under the terms of the MIT license.

🔼 Back to top