Actian Data Platform, Actian X, Ingres, and Vector dialect for SQLAlchemy CPython 3.x (and 2.7).

For more information about SQLAlchemy see:

• https://github.com/sqlalchemy/sqlalchemy
• https://www.sqlalchemy.org/
• https://pypi.org/project/SQLAlchemy/

The Ingres dialect was originally developed to work with SQLAlchemy versions 0.6 and Ingres 9.2. The current code has recently been thoroughly tested with these versions:

• SQLAlchemy 1.4.51 and 2.0.30
• Actian Data Platform, Ingres 11.x & 12.x, Vector 5.x & 6.x - via ODBC

It is important to be aware of which version of SQLAlchemy is installed. Version pinning provides the ability to install the desired version explicitly. Version pinning examples:

    python -m pip install 'sqlalchemy < 2.0.30'
    python -m pip install sqlalchemy==1.4.51

Jython/JDBC support is currently untested, as the current code relies on zxjdbc it is not recommended this be used (see https://hg.sr.ht/~clach04/jyjdbc for as an alternative that includes full Decimal datatype support).

Known to work with:

• https://github.com/cloudera/hue
• https://github.com/apache/superset (see https://github.com/clach04/incubator-superset/tree/vector)
• https://github.com/catherinedevlin/ipython-sql / Jupyter/IPython notebooks (see catherinedevlin/ipython-sql#196 - or use %config SqlMagic.autocommit=False
  • Until ipython-sql 0.4.1 is released, to avoid workaround issue; pip install git+https://github.com/catherinedevlin/ipython-sql.git
• https://github.com/wireservice/csvkit (see wireservice/agate-sql#36)

---

• Quickstart
  • Install from PyPi
  • Install latest from GitHub, without a source code checkout
  • Install latest from GitHub, with a source code checkout
• Development instructions
  • Quick python test
  • Troubleshooting
  • Running SA test suite

---

Quickstart

TL;DR

Doc ref: pip install

Install from PyPi, along with required packages

python -m pip install "sqlalchemy-ingres[all]"

Install latest from GitHub, without a source code checkout

python -m pip install -e git+https://github.com/ActianCorp/sqlalchemy-ingres.git#egg=sqlalchemy-ingres

Alternatively, for a named_branch:

python -m pip install -e git+https://github.com/ActianCorp/sqlalchemy-ingres.git@named_branch#egg=sqlalchemy-ingres

Install latest from GitHub, with a source code checkout

Install/setup:

python -m pip install pyodbc sqlalchemy
git clone https://github.com/ActianCorp/sqlalchemy-ingres.git
cd sqlalchemy-ingres
python -m pip install -e .

Demo Sample / Sanity check:

Linux/Unix

export SQLALCHEMY_INGRES_ODBC_DRIVER_NAME=INGRES Y1
# or what ever the ODBC Driver name is; Actian, Ingres, etc.
# Only needed if program/environment is unable to identify the correct driver.

Windows:

SET SQLALCHEMY_INGRES_ODBC_DRIVER_NAME=INGRES Y1
REM or what ever the ODBC Driver name is; Actian, Ingres, etc.
REM Only needed if program/environment is unable to identify the correct driver.

NOTE ODBC Driver should be the same bitness as the Python interpreter. That is:

• for 64-bit Python ensure a 64-bit ODBC driver is available.
• For 32-bit Python ensure a 32-bit ODBC driver is available.

Assuming local DBMS, python session:

Python 3.7.3 (v3.7.3:ef4ec6ed12, Mar 25 2019, 22:22:05) [MSC v.1916 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import sqlalchemy
>>> engine = sqlalchemy.create_engine('ingres:///iidbdb')  # local DBMS
>>> connection = engine.connect()
>>> for row in connection.execute(sqlalchemy.text('SELECT count(*) FROM iidatabase')):
...     print(row)
...
(33,)

Development instructions

Right now this is for dev purposes so install SQLAlchemy as per normal, for example:

python -m pip install sqlalchemy

or for dev testing and modifying/running tests:

git clone -b rel_1_4_42 https://github.com/sqlalchemy/sqlalchemy.git
cd sqlalchemy
#pip install -e .
python -m pip install -e .   # https://adamj.eu/tech/2020/02/25/use-python-m-pip-everywhere/

Ingres dialect tested with pyodbc and pypyodbc (pypyodbc useful for debugging, to see what SQLAlchemy is calling with):

pip install pyodbc

Download Ingres dialect for SQLAlchemy:

git clone https://github.com/ActianCorp/sqlalchemy-ingres.git

Setup for dev use:

cd sqlalchemy-ingres
python -m pip install -e .

Demo/Test:

Ensure Ingres ODBC driver is available, Ingres sqlalchemy defaults to using "Ingres" ODBC driver, ensure this is the same bit-age as the Python interpreter. For example, for 64-bit Python, ensure ODBC Driver called "Ingres" is also 64-bit.

ODBC Driver name can be overridden via environment variable SQLALCHEMY_INGRES_ODBC_DRIVER_NAME, for example:

Windows 64-bit:

set SQLALCHEMY_INGRES_ODBC_DRIVER_NAME=Ingres CS
REM Only needed if program/environment is unable to identify the correct driver.

Windows 32-bit:

set SQLALCHEMY_INGRES_ODBC_DRIVER_NAME=Ingres CR
REM Only needed if program/environment is unable to identify the correct driver.

Under Linux/Unix check ODBC settings and if using UnixODBC, check how wide-char support was built, recommendation for out-of-box Linux distributions:

# Most Linux distros build UnixODBC with non-default build options
export II_ODBC_WCHAR_SIZE=2

# set variables for ODBC config
export ODBCSYSINI=$II_SYSTEM/ingres/files
export ODBCINI=$II_SYSTEM/ingres/files/odbc.ini

Quick python test

import sys
import sqlalchemy
#import sqlalchemy_ingres

print('Python %s on %s' % (sys.version, sys.platform))
print('SQLAlchemy %r' % sqlalchemy.__version__)

con_str = 'ingres:///demodb'  # local demodb
#con_str = 'ingres://dbuser:PASSWORD@HOSTNAME:27832/db'  # remote database called "db"
print(con_str)

# If the next line is uncommented, need to also uncomment: import sqlalchemy_ingres
#print(sqlalchemy_ingres.base.dialect().create_connect_args(url=sqlalchemy.engine.make_url(con_str)))

engine = sqlalchemy.create_engine(con_str)
connection = engine.connect()

query = 'SELECT * FROM iidbconstants'
for row in connection.execute(sqlalchemy.text(query)):
    print(row)

Troubleshooting

Getting error:

sqlalchemy.exc.DatabaseError: (pypyodbc.DatabaseError) ('08004', '[08004] [Actian][Ingres ODBC Driver][INGRES]Requested association partner is unavailable')

1. DBMS may not be running or accesible (e.g. network error).

2. Could be using a Driver name that is not available (or the wrong number of bits, e.g. 32-bit versus 64-bit or vice-versa), or wrong environment (e.g. multiple DBMS/client installations). Solution, make use of environment variable SQLALCHEMY_INGRES_ODBC_DRIVER_NAME either in the environment or in Python code, e.g:

   import os; os.environ['SQLALCHEMY_INGRES_ODBC_DRIVER_NAME'] = 'Ingres X2'  # Etc. where X2 is the installation id (output from, "ingprenv II_INSTALLATION")

Running SA test suite

NOTE below is for Python 2.7 and 3.4, can remove version pin for current python. Mock appears to be a dependency that is not pulled in for py2.7

python -m pip  install --upgrade pip
python -m pip  install tox "pytest==4.6"
python -m pip  install mock

Setup test config

$ cat test.cfg
# test.cfg file
# see README.unittests.rst
#       pytest --db sqlite_file
[db]
sqlite_file=sqlite:///querytest.sqlite3

# local
ingres_odbc=ingres:///sa

Code change needed to SA:

diff --git a/test/requirements.py b/test/requirements.py
index cf9168f5a..fcc4f37a0 100644
--- a/test/requirements.py
+++ b/test/requirements.py
@@ -394,6 +394,9 @@ class DefaultRequirements(SuiteRequirements):
         elif against(config, "oracle"):
             default = "READ COMMITTED"
             levels.add("AUTOCOMMIT")
+        elif against(config, "ingres"):
+            default = "READ COMMITTED"
+            levels.add("AUTOCOMMIT")  # probably needed, not sure what this is though - assuming tests are not commiting and expecting autocommit semantics
         else:
             raise NotImplementedError()

Run (all) tests:

pytest --db ingres_odbc --junit-xml=all_results_junit.xml --maxfail=12000

Execution Options

Index Reflection

The default Ingres dialect behavior for index reflection is to return only user-defined indexes.

To inspect all indexes, including those generated by the DBMS, set the connection execution option inspect_indexes="ALL".

Example:

connection.execution_options(inspect_indexes="ALL")
i = sqlalchemy.inspect(connection)
print(i.get_indexes("employee_table"))

Documentation reference iiindexes catalog

Known Issues and Limitations

Apache Superset issue 27427

The Apache Superset SQL parser is able to recognize and handle row limiting clauses that use keywords LIMIT and TOP but does not handle the FETCH FIRST clause, which is used by some databases, including Ingres and PostgreSQL.

If a FETCH FIRST clause is used in a SQL statement and the Superset config parameter SQL_MAX_ROW is set to a value > 0, the Superset parser appends a LIMIT clause to the SQL statement, making it syntactically invalid.

The errant behavior can be seen when working with SQL statements and table metadata in Superset SQL Lab and may occur in other Superset operations as well.

It is important to note that this is a problem with Apache Superset and not with the SQLAlchemy-Ingres connector.

Reflection of constraint metadata does not include attributes referential_actions and enforce_option

Actian databases provide optional clauses for specifying referential_actions and enforce_option when defining table-level or column-level constraints.

Clauses from doc:

referential_actions

• ON UPDATE CASCADE
• ON UPDATE {RESTRICT | NO ACTION }
• ON DELETE CASCADE
• ON DELETE {RESTRICT | NO ACTION }
• NO ACTION
• RESTRICT
• SET NULL

enforce_option

• NOT ENFORCED
• ENFORCED

When reflecting objects with constraints that are defined having these optional clauses, the returned metadata does not include information on these attributes.