PyPika 0.27.2

A SQL query builder API for Python

Abstract

What is PyPika?

PyPika is a Python API for building SQL queries. The motivation behind PyPika is to provide a simple interface for building SQL queries without limiting the flexibility of handwritten SQL. Designed with data analysis in mind, PyPika leverages the builder design pattern to construct queries to avoid messy string formatting and concatenation. It is also easily extended to take full advantage of specific features of SQL database vendors.

What are the design goals for PyPika?

PyPika is a fast, expressive and flexible way to replace handwritten SQL (or even ORM for the courageous souls amongst you). Validation of SQL correctness is not an explicit goal of PyPika. With such a large number of SQL database vendors providing a robust validation of input data is difficult. Instead you are encouraged to check inputs you provide to PyPika or appropriately handle errors raised from your SQL database - just as you would have if you were writing SQL yourself.

Read the docs: http://pypika.readthedocs.io/en/latest/

Installation

PyPika supports python 2.7 and 3.3+. It may also work on pypy, cython, and jython, but is not being tested for these versions.

To install PyPika run the following command:

pip install pypika

Tutorial

The main classes in pypika are pypika.Query, pypika.Table, and pypika.Field.

from pypika import Query, Table, Field

Selecting Data

The entry point for building queries is pypika.Query. In order to select columns from a table, the table must first be added to the query. For simple queries with only one table, tables and columns can be references using strings. For more sophisticated queries a pypika.Table must be used.

q = Query.from_('customers').select('id', 'fname', 'lname', 'phone')

To convert the query into raw SQL, it can be cast to a string.

str(q)

Alternatively, you can use the Query.get_sql() function:

q.get_sql()

Using pypika.Table

customers = Table('customers')
q = Query.from_(customers).select(customers.id, customers.fname, customers.lname, customers.phone)

Both of the above examples result in the following SQL:

SELECT id,fname,lname,phone FROM customers

Results can be ordered by using the following syntax:

from pypika import Order
Query.from_('customers').select('id', 'fname', 'lname', 'phone').orderby('id', order=Order.desc)

This results in the following SQL:

SELECT "id","fname","lname","phone" FROM "customers" ORDER BY "id" DESC

Arithmetic

Arithmetic expressions can also be constructed using pypika. Operators such as +, -, *, and / are implemented by pypika.Field which can be used simply with a pypika.Table or directly.

from pypika import Field

q = Query.from_('account').select(
    Field('revenue') - Field('cost')
)

SELECT revenue-cost FROM accounts

Using pypika.Table

accounts = Table('accounts')
q = Query.from_(accounts).select(
    accounts.revenue - accounts.cost
)

SELECT revenue-cost FROM accounts

An alias can also be used for fields and expressions.

q = Query.from_(accounts).select(
    (accounts.revenue - accounts.cost).as_('profit')
)

SELECT revenue-cost profit FROM accounts

More arithmetic examples

table = Table('table')
q = Query.from_(table).select(
    table.foo + table.bar,
    table.foo - table.bar,
    table.foo * table.bar,
    table.foo / table.bar,
    (table.foo+table.bar) / table.fiz,
)

SELECT foo+bar,foo-bar,foo*bar,foo/bar,(foo+bar)/fiz FROM table

Filtering

Queries can be filtered with pypika.Criterion by using equality or inequality operators

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id, customers.fname, customers.lname, customers.phone
).where(
    customers.lname == 'Mustermann'
)

SELECT id,fname,lname,phone FROM customers WHERE lname='Mustermann'

Query methods such as select, where, groupby, and orderby can be called multiple times. Multiple calls to the where method will add additional conditions as

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id, customers.fname, customers.lname, customers.phone
).where(
    customers.fname == 'Max'
).where(
    customers.lname == 'Mustermann'
)

SELECT id,fname,lname,phone FROM customers WHERE fname='Max' AND lname='Mustermann'

Filters such as IN and BETWEEN are also supported

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id,customers.fname
).where(
    customers.age[18:65] & customers.status.isin(['new', 'active'])
)

SELECT id,fname FROM customers WHERE age BETWEEN 18 AND 65 AND status IN ('new','active')

Filtering with complex criteria can be created using boolean symbols &, |, and ^.

AND

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id, customers.fname, customers.lname, customers.phone
).where(
    (customers.age >= 18) & (customers.lname == 'Mustermann')
)

SELECT id,fname,lname,phone FROM customers WHERE age>=18 AND lname='Mustermann'

OR

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id, customers.fname, customers.lname, customers.phone
).where(
    (customers.age >= 18) | (customers.lname == 'Mustermann')
)

SELECT id,fname,lname,phone FROM customers WHERE age>=18 OR lname='Mustermann'

XOR

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id, customers.fname, customers.lname, customers.phone
).where(
    (customers.age >= 18) ^ customers.is_registered
)

SELECT id,fname,lname,phone FROM customers WHERE age>=18 XOR is_registered

Convenience Methods

In the Criterion class, there are the static methods any and all that allow building chains AND and OR expressions with a list of terms.

from pypika import Criterion

customers = Table('customers')
q = Query.from_(customers).select(
    customers.id,
    customers.fname
).where(
    Criterion.all([
        customers.is_registered,
        customers.age >= 18,
        customers.lname == "Jones",
    ])
)

SELECT id,fname FROM customers WHERE is_registered AND age>=18 AND lname = "Jones"

Grouping and Aggregating

Grouping allows for aggregated results and works similar to SELECT clauses.

from pypika import functions as fn

customers = Table('customers')
q = Query.from_(customers).where(
    customers.age >= 18
).groupby(
    customers.id
).select(
    customers.id, fn.Sum(customers.revenue)
)

SELECT id,SUM(revenue) FROM customers WHERE age>=18 GROUP BY id ORDER BY id ASC

After adding a GROUP BY clause to a query, the HAVING clause becomes available. The method Query.having() takes a Criterion parameter similar to the method Query.where().

from pypika import functions as fn

payments = Table('payments')
q = Query.from_(payments).where(
    payments.transacted[date(2015, 1, 1):date(2016, 1, 1)]
).groupby(
    payments.customer_id
).having(
    fn.Sum(payments.total) >= 1000
).select(
    payments.customer_id, fn.Sum(payments.total)
)

SELECT customer_id,SUM(total) FROM payments
WHERE transacted BETWEEN '2015-01-01' AND '2016-01-01'
GROUP BY customer_id HAVING SUM(total)>=1000

Joining Tables and Subqueries

Tables and subqueries can be joined to any query using the Query.join() method. Joins can be performed with either a USING or ON clauses. The USING clause can be used when both tables/subqueries contain the same field and the ON clause can be used with a criterion. To perform a join, ...join() can be chained but then must be followed immediately by ...on(<criterion>) or ...using(*field).

Example of a join using ON

history, customers = Tables('history', 'customers')
q = Query.from_(history).join(
    customers
).on(
    history.customer_id == customers.id
).select(
    history.star
).where(
    customers.id == 5
)

SELECT "history".* FROM "history" JOIN "customers" ON "history"."customer_id"="customers"."id" WHERE "customers"."id"=5

As a shortcut, the Query.join().on_field() function is provided for joining the (first) table in the FROM clause with the joined table when the field name(s) are the same in both tables.

Example of a join using ON

history, customers = Tables('history', 'customers')
q = Query.from_(history).join(
    customers
).on_field(
    'customer_id', 'group'
).select(
    history.star
).where(
    customers.group == 'A'
)

SELECT "history".* FROM "history" JOIN "customers" ON "history"."customer_id"="customers"."customer_id" AND "history"."group"="customers"."group" WHERE "customers"."group"='A'

Example of a join using USING

history, customers = Tables('history', 'customers')
q = Query.from_(history).join(
    customers
).on(
    'customer_id'
).select(
    history.star
).where(
    customers.id == 5
)

SELECT "history".* FROM "history" JOIN "customers" USING "customer_id" WHERE "customers"."id"=5

Example of a correlated subquery in the SELECT

history, customers = Tables('history', 'customers')
last_purchase_at = Query.from_(history).select(
    history.purchase_at
).where(history.customer_id==customers.customer_id).orderby(
    history.purchase_at, order=Order.desc
).limit(1)
q = Query.from_(customers).select(
    customers.id, last_purchase_at._as('last_purchase_at')
)

SELECT
  "id",
  (SELECT "history"."purchase_at"
   FROM "history"
   WHERE "history"."customer_id" = "customers"."customer_id"
   ORDER BY "history"."purchase_at" DESC
   LIMIT 1) "last_purchase_at"
FROM "customers"

Unions

Both UNION and UNION ALL are supported. UNION DISTINCT is synonomous with “UNION`` so and PyPika does not provide a separate function for it. Unions require that queries have the same number of SELECT clauses so trying to cast a unioned query to string with through a UnionException if the column sizes are mismatched.

To create a union query, use either the Query.union() method or + operator with two query instances. For a union all, use Query.union_all() or the * operator.

provider_a, provider_b = Tables('provider_a', 'provider_b')
q = Query.from_(provider_a).select(
    provider_a.created_time, provider_a.foo, provider_a.bar
) + Query.from_(provider_b).select(
    provider_b.created_time, provider_b.fiz, provider_b.buz
)

SELECT "created_time","foo","bar" FROM "provider_a" UNION SELECT "created_time","fiz","buz" FROM "provider_b"

Date, Time, and Intervals

Using pypika.Interval, queries can be constructed with date arithmetic. Any combination of intervals can be used except for weeks and quarters, which must be used separately and will ignore any other values if selected.

from pypika import functions as fn

fruits = Tables('fruits')
q = Query.from_(fruits) \
    .select(fruits.id, fruits.name) \
    .where(fruits.harvest_date + Interval(months=1) < fn.Now())

SELECT id,name FROM fruits WHERE harvest_date+INTERVAL 1 MONTH<NOW()

Tuples

Tuples are supported through the class pypika.Tuple but also through the native python tuple wherever possible. Tuples can be used with pypika.Criterion in WHERE clauses for pairwise comparisons.

from pypika import Query, Tuple

q = Query.from_(self.table_abc) \
    .select(self.table_abc.foo, self.table_abc.bar) \
    .where(Tuple(self.table_abc.foo, self.table_abc.bar) == Tuple(1, 2))

SELECT "foo","bar" FROM "abc" WHERE ("foo","bar")=(1,2)

Using pypika.Tuple on both sides of the comparison is redundant and PyPika supports native python tuples.

from pypika import Query, Tuple

q = Query.from_(self.table_abc) \
    .select(self.table_abc.foo, self.table_abc.bar) \
    .where(Tuple(self.table_abc.foo, self.table_abc.bar) == (1, 2))

SELECT "foo","bar" FROM "abc" WHERE ("foo","bar")=(1,2)

Tuples can be used in IN clauses.

Query.from_(self.table_abc) \
        .select(self.table_abc.foo, self.table_abc.bar) \
        .where(Tuple(self.table_abc.foo, self.table_abc.bar).isin([(1, 1), (2, 2), (3, 3)]))

SELECT "foo","bar" FROM "abc" WHERE ("foo","bar") IN ((1,1),(2,2),(3,3))

Strings Functions

There are several string operations and function wrappers included in PyPika. Function wrappers can be found in the pypika.functions package. In addition, LIKE and REGEX queries are supported as well.

from pypika import functions as fn

customers = Tables('customers')
q = Query.from_(customers).select(
    customers.id,
    customers.fname,
    customers.lname,
).where(
    customers.lname.like('Mc%')
)

SELECT id,fname,lname FROM customers WHERE lname LIKE 'Mc%'

from pypika import functions as fn

customers = Tables('customers')
q = Query.from_(customers).select(
    customers.id,
    customers.fname,
    customers.lname,
).where(
    customers.lname.regex(r'^[abc][a-zA-Z]+&')
)

SELECT id,fname,lname FROM customers WHERE lname REGEX '^[abc][a-zA-Z]+&';

from pypika import functions as fn

customers = Tables('customers')
q = Query.from_(customers).select(
    customers.id,
    fn.Concat(customers.fname, ' ', customers.lname).as_('full_name'),
)

SELECT id,CONCAT(fname, ' ', lname) full_name FROM customers

Case Statements

Case statements allow fow a number of conditions to be checked sequentially and return a value for the first condition met or otherwise a default value. The Case object can be used to chain conditions together along with their output using the when method and to set the default value using else_.

from pypika import Case, functions as fn

customers = Tables('customers')
q = Query.from_(customers).select(
    customers.id,
    Case()
       .when(customers.fname == "Tom", "It was Tom")
       .when(customers.fname == "John", "It was John")
       .else_("It was someone else.").as_('who_was_it')
)

SELECT "id",CASE WHEN "fname"='Tom' THEN 'It was Tom' WHEN "fname"='John' THEN 'It was John' ELSE 'It was someone else.' END "who_was_it" FROM "customers"

Inserting Data

Data can be inserted into tables either by providing the values in the query or by selecting them through another query.

By default, data can be inserted by providing values for all columns in the order that they are defined in the table.

Insert with values

customers = Table('customers')

q = Query.into(customers).insert(1, 'Jane', 'Doe', '[e-mail]')

INSERT INTO customers VALUES (1,'Jane','Doe','[e-mail]')

Multiple rows of data can be inserted either by chaining the insert function or passing multiple tuples as args.

customers = Table('customers')

q = Query.into(customers).insert(1, 'Jane', 'Doe', '[e-mail]').insert(2, 'John', 'Doe', '[e-mail]')

customers = Table('customers')

q = Query.into(customers).insert((1, 'Jane', 'Doe', '[e-mail]'),
                                 (2, 'John', 'Doe', '[e-mail]'))

Insert with on Duplicate Key Update

customers = Table('customers')

q = Query.into(customers)\
    .insert(1, 'Jane', 'Doe', '[e-mail]')\
    .on_duplicate_key_update(customers.email, Values(customers.email))

INSERT INTO customers VALUES (1,'Jane','Doe','[e-mail]') ON DUPLICATE KEY UPDATE `email`=VALUES(`email`)

.on_duplicate_key_update works similar to .set for updating rows, additionally it provides the Values wrapper to update to the value specified in the INSERT clause.

Insert from a SELECT Sub-query

INSERT INTO customers VALUES (1,'Jane','Doe','[e-mail]'),(2,'John','Doe','[e-mail]')

To specify the columns and the order, use the columns function.

customers = Table('customers')

q = Query.into(customers).columns('id', 'fname', 'lname').insert(1, 'Jane', 'Doe')

INSERT INTO customers (id,fname,lname) VALUES (1,'Jane','Doe','[e-mail]')

Inserting data with a query works the same as querying data with the additional call to the into method in the builder chain.

customers, customers_backup = Tables('customers', 'customers_backup')

q = Query.into(customers_backup).from_(customers).select('*')

INSERT INTO customers_backup SELECT * FROM customers

customers, customers_backup = Tables('customers', 'customers_backup')

q = Query.into(customers_backup).columns('id', 'fname', 'lname')
    .from_(customers).select(customers.id, customers.fname, customers.lname)

INSERT INTO customers_backup SELECT "id", "fname", "lname" FROM customers

The syntax for joining tables is the same as when selecting data

customers, orders, orders_backup = Tables('customers', 'orders', 'orders_backup')

q = Query.into(orders_backup).columns('id', 'address', 'customer_fname', 'customer_lname')
    .from_(customers)
    .join(orders).on(orders.customer_id == customers.id)
    .select(orders.id, customers.fname, customers.lname)

INSERT INTO "orders_backup" ("id","address","customer_fname","customer_lname")
SELECT "orders"."id","customers"."fname","customers"."lname" FROM "customers"
JOIN "orders" ON "orders"."customer_id"="customers"."id"

Updating Data

PyPika allows update queries to be constructed with or without where clauses.

customers = Table('customers')

Query.update(customers).set(customers.last_login, '2017-01-01 10:00:00')

Query.update(customers).set(customers.lname, smith).where(customers.id == 10)

UPDATE "customers" SET "last_login"='2017-01-01 10:00:00'

UPDATE "customers" SET "lname"='smith' WHERE "id"=10

The syntax for joining tables is the same as when selecting data

customers, profiles = Tables('customers', 'profiles')

Query.update(customers)
     .join(profiles).on(profiles.customer_id == customers.id)
     .set(customers.lname, profiles.lname)

UPDATE "customers"
JOIN "profiles" ON "profiles"."customer_id"="customers"."id"
SET "customers"."lname"="profiles"."lname"

Parametrized Queries

PyPika allows you to use Parameter(str) term as a placeholder for parametrized queries.

customers = Table('customers')

q = Query.into(customers).columns('id', 'fname', 'lname')
    .insert(Parameter(':1'), Parameter(':2'), Parameter(':3'))

INSERT INTO customers (id,fname,lname) VALUES (:1,:2,:3)

This allows you to build prepared statements, and/or avoid SQL-injection related risks.

Due to the mix of syntax for parameters, depending on connector/driver, it is required that you specify the parameter token explicitly.

An example of some common SQL parameter styles used in Python drivers are:

PostgreSQL:

$number OR %s + :name (depending on driver)

MySQL:

%s

SQLite:

?

Vertica:

:name

Oracle:

:number + :name

MSSQL:

%(name)s OR :name + :number (depending on driver)

You can find out what parameter style is needed for DBAPI compliant drivers here: https://www.python.org/dev/peps/pep-0249/#paramstyle or in the DB driver documentation.

License

Copyright 2016 KAYAK Germany, GmbH

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Crafted with ♥ in Berlin.