1. Chris McDonough
  2. htsql


htsql / INSTALL

Installation Instructions

The following installation instructions were tested under a fresh installation
of Ubuntu 10.04 Server, but they could be easily adapted to other Linux
distributions and package managers.

TL;DR version

1. Install Python and required Python modules::

    # apt-get install python
    # apt-get install python-setuptools python-simplejson python-pyyaml python-psycopg2

2. Install Mercurial and download HTSQL source code::

    # apt-get install mercurial
    # hg clone http://bitbucket.org/prometheus/htsql

3. Build and install HTSQL::

    # cd htsql
    # make build
    # make install

   This should create a ``htsql-ctl`` script.  Run::

    $ htsql-ctl help

   for general help and list of commands.  Run::


   to start an HTSQL server on the address `HOST:PORT` against the specified
   database.  In particular, to run the HTSQL server against a PostgreSQL
   database deployed on the same machine with credentials of the current
   user, run::

    $ htsql-ctl server pgsql:///DATABASE

Installing prerequisites

HTSQL requires Python 2.5 or newer, but does not support Python 3 yet.
Python 2.6 is the recommended version.  In some distributions, Python is
already installed; if not, you could install it by running::

    # apt-get install python

HTSQL depends on the following external Python packages:

 * `setuptools` (0.6c9 or newer);
 * `simplejson` (2.0 or newer);
 * `PyYAML` (3.07 or newer, compiled with LibYAML bindings);
 * `psycopg2` (2.0.10 or newer, earlier versions are known to segfault).

To install the packages, run::

    # apt-get install python-setuptools python-simplejson python-pyyaml python-psycopg2

Alternatively, you could install these Python modules from sources using the
`easy_install` script from `setuptools`.  To do that, you need to install
header files for Python, PostgreSQL and LibYAML::

    # apt-get install build-essential python-dev libpq-dev libyaml-dev
    # apt-get install python-setuptools
    # easy_install simplejson
    # easy_install PyYAML
    # easy_install psycopg2

This method allows you to choose the directory where the external Python
packages will be installed.  By default, `easy_install` installs packages
under ``/usr/local`` directory.  To install Python packages under your home
directory, create a file ``.pydistutils.cfg`` in your home directory with
the following content::


and set the environment variable ``PYTHONUSERBASE``::


Then running `easy_install` will install Python libraries to
``~/lib/python2.6/site-packages`` (when running under Python 2.6) and Python
scripts to ``~/bin``.

For more details on customizing the location of external Python modules, see
`distutils` and `setuptools` documentation.

Installing HTSQL

Once HTSQL is officially released, you will be able to install it using the
`easy_install` script::

    # easy_install HTSQL

That will download, build and install the latest released version of HTSQL.

Alternatively, you can install HTSQL from the Mercurial repository at
BitBucket.  You need a Mercurial client::

    # apt-get install mercurial

Then download HTSQL sources::

    # hg clone http://bitbucket.org/prometheus/htsql

To build and install HTSQL, run::

    # cd htsql
    # make build
    # make install

To install HTSQL in the development mode, run::

    # make develop

When HTSQL is installed in the development mode, any changes in the source
files are reflected immediately without having to reinstall it again.

HTSQL comes with a comprehensive test suite.  Running the regression tests
requires a PostgreSQL server instance.  By default, the regression tests
assume that the database server is installed locally and the current user
has administrative permissions.  To install a local PostgreSQL server, run::

    # apt-get install postgresql

To add a database user with the same name as your login name, run::

    # su - postgres -c "createuser -s $USER"

If the host is a single user machine, it is often convenient to allow
any user on the system connect to the database under any database user
name.  To do it, open the file ``/etc/postgresql/8.4/main/pg_hba.conf``
(replace the version number with the actual version), find the lines::

    # "local" is for Unix domain socket connections only
    local   all         all                               ident

and replace them with::

    # "local" is for Unix domain socket connections only
    local   all         all                               trust

Reload the server configuration::

    # service postgresql-8.4 reload

Alternatively, if you already have a PostgreSQL server installed somewhere,
you can specify the address and connection parameters explicitly.  Copy the
file ``Makefile.env.sample`` to ``Makefile.env``, open the latter, and edit
the values of variables: ``PGSQL_ADMIN_USERNAME``, ``PGSQL_ADMIN_PASSWORD``,
``PGSQL_HOST``, ``PGSQL_PORT``.  They should contain the credentials of
an administrative user and the address of the server respectively.

To run HTSQL regression tests, run::

    # make test

Running regression tests creates a database user called ``htsql_regress``
with the password ``secret``, and a database called ``htsql_regress``.  Feel
free to use this database for playing with HTSQL.

To remove any database users and databases deployed by the regression tests,

    # make cleanup

To build the documentation that comes with HTSQL, run

    # make doc

Note that this requires Sphinx 1.0, which is not yet released.

Running HTSQL

If HTSQL is installed successfully, you should be able to run the
`htsql-ctl` script::

    $ htsql-ctl

The script has a number of subcommands called *routines*.  In general, the
command line has the form::

    htsql-ctl <routine> [options] [arguments]

where ``<routine>`` is the routine name, ``options`` is any routine options
in short (``-X``) or long (``--option-name``) form, and ``arguments`` is the
routine arguments.  Run::

    $ htsql-ctl help

to get a list of routines and::

    $ htsql-ctl help <routine>

to describe a specific routine.

To start an HTSQL server, run


Here `ENGINE` is either ``pgsql`` or ``sqlite``; `USERNAME:PASSWORD` are
used for authentication; `HOST:PORT` is the address of the database server;
and `DATABASE` is the name of the database to connect.  All parameters
except for `ENGINE` and `DATABASE` are optional.  For instance::

    $ htsql-ctl server pgsql:///htsql_regress

will start the HTSQL server against the HTSQL regression database (provided
it is deployed by running the regression tests).

By default the server is listening on ``localhost:8080``.  To specify a
different address of the HTSQL server, use optional arguments `HOST` and


For more help on the ``server`` routine, run::

    $ htsql-ctl help server

The script also allows you to run HTSQL queries from the console using the
``shell`` routine.  To start the shell, run::


This will display the command prompt where you could type and execute HTSQL
queries.  For more details, run

    $ htsql-ctl help shell

or type ``help`` in the shell.

.. TODO: Installing on MS Windows

Have fun and enjoy!