=encoding UTF-8

=head1 Name

sqitchtutorial - A tutorial introduction to Sqitch change management on PostgreSQL

=head1 Synopsis

sqitch *

=head1 Description

This tutorial explains how to create a sqitch-enabled PostgreSQL project, use

a VCS for deployment planning, and work with other developers to make sure

changes remain in sync and in the proper order.

We'll start by creating a new project from scratch, a fictional antisocial

networking site called Flipr. All examples use L<Git|https://git-scm.com/> as

the VCS and L<PostgreSQL|https://www.postgresql.org/> as the storage engine,

though L<YugabyteDB|https://www.yugabyte.com/yugabytedb/> and

L<CockroachDB|https://www.cockroachlabs.com/product/> should work just as well.

For the most part you can substitute other VCSes and database engines in

the examples as appropriate.

If you'd like to manage an SQLite database, see L<sqitchtutorial-sqlite>.

If you'd like to manage an Oracle database, see L<sqitchtutorial-oracle>.

If you'd like to manage a MySQL database, see L<sqitchtutorial-mysql>.

If you'd like to manage a Firebird database, see L<sqitchtutorial-firebird>.

If you'd like to manage a Vertica database, see L<sqitchtutorial-vertica>.

If you'd like to manage an Exasol database, see L<sqitchtutorial-exasol>.

If you'd like to manage a Snowflake database, see L<sqitchtutorial-snowflake>.

=head1 Starting a New Project

Usually the first thing to do when starting a new project is to create a

source code repository. So let's do that with Git:

> mkdir flipr

> cd flipr

> git init .

Initialized empty Git repository in /flipr/.git/

> touch README.md

> git add .

> git commit -am 'Initialize project, add README.'

If you're a Git user and want to follow along the history, the repository

used in these examples is L<on GitHub|https://github.com/sqitchers/sqitch-intro>.

Now that we have a repository, let's get started with Sqitch. Every Sqitch

project must have a name associated with it, and, optionally, a unique URI. We

recommend including the URI, as it increases the uniqueness of object

identifiers internally, and will prevent the deployment of a different project

with the same name. So let's specify one when we initialize Sqitch:

> sqitch init flipr --uri https://github.com/sqitchers/sqitch-intro/ --engine pg

Created sqitch.conf

Created sqitch.plan

Created deploy/

Created revert/

Created verify/

Let's have a look at F<sqitch.conf>:

> cat sqitch.conf

[core]

engine = pg

# plan_file = sqitch.plan

# top_dir = .

# [engine "pg"]

# target = db:pg:

# registry = sqitch

# client = psql

Good, it picked up on the fact that we're creating changes for the PostgreSQL

engine, thanks to the C<--engine pg> option, and saved it to the file.

Furthermore, it wrote a commented-out C<[engine "pg"]> section with all the

available PostgreSQL engine-specific settings commented out and ready to be

edited as appropriate. This configuration will also fork for

L<YugabyteDB|https://www.yugabyte.com/yugabytedb/> databases, too. For

L<CockroachDB|https://www.cockroachlabs.com/product/>, use

C<--engine cockroach>, instead, and replace any instance of C<pg> with

C<cockroach> in the rest of this document.

By default, Sqitch will read F<sqitch.conf> in the current directory for

settings. But it will also read F<~/.sqitch/sqitch.conf> for user-specific

settings. Since PostgreSQL's C<psql> client is not in the path on my system,

let's go ahead and tell it where to find the client on our computer (don't

bother if you're using the

L<Docker image|https://hub.docker.com/r/sqitch/sqitch/> because it uses the

client inside the container, not on your host machine):

> sqitch config --user engine.pg.client /opt/local/pgsql/bin/psql

And let's also tell it who we are, since this data will be used in all

of our projects:

> sqitch config --user user.name 'Marge N. O’Vera'

> sqitch config --user user.email 'marge@example.com'

Have a look at F<~/.sqitch/sqitch.conf> and you'll see this:

> cat ~/.sqitch/sqitch.conf

[engine "pg"]

client = /opt/local/pgsql/bin/psql

[user]

name = Marge N. O’Vera

email = marge@example.com

Which means that Sqitch should be able to find C<psql> for any project, and

that it will always properly identify us when planning and committing changes.

Back to the repository. Have a look at the plan file, F<sqitch.plan>:

> cat sqitch.plan

%syntax-version=1.0.0

%project=flipr

%uri=https://github.com/sqitchers/sqitch-intro/

Note that it has picked up on the name and URI of the app we're building.

Sqitch uses this data to manage cross-project dependencies. The

C<%syntax-version> pragma is always set by Sqitch, so that it always knows how

to parse the plan, even if the format changes in the future.

Let's commit these changes and start creating the database changes.

> git add .

> git commit -am 'Initialize Sqitch configuration.'

[main 85e8d7c] Initialize Sqitch configuration.

2 files changed, 19 insertions(+)

create mode 100644 sqitch.conf

create mode 100644 sqitch.plan

=head1 Our First Change

First, our project will need a schema. This creates a nice namespace for all

of the objects that will be part of the flipr app. Run this command:

> sqitch add appschema -n 'Add schema for all flipr objects.'

Created deploy/appschema.sql

Created revert/appschema.sql

Created verify/appschema.sql

Added "appschema" to sqitch.plan

The L<C<add>|sqitch-add> command adds a database change to the plan and writes

deploy, revert, and verify scripts that represent the change. Now we edit

these files. The C<deploy> script's job is to create the schema. So we add

this to F<deploy/appschema.sql>:

CREATE SCHEMA flipr;

The C<revert> script's job is to precisely revert the change to the deploy

script, so we add this to F<revert/appschema.sql>:

DROP SCHEMA flipr;

Now we can try deploying this change. First, we need to create a database

to deploy to:

> createdb flipr_test

Now we tell Sqitch where to send the change via a

L<database URI|https://github.com/libwww-perl/uri-db/>. The examples here

use C<db:pg:flipr_test>, assuming a passwordless connection on a local

socket. If you need to specify a username and hostname, it would be more

like C<db:pg://postgres@localhost/flipr_test>. For YugabyteDB, you might

also need the port, as in C<db:pg://postgres@localhost:5433/flipr_test>.

For CockroachDB, use C<db:cockroach:>, as in

C<db:cockroach://root@localhost:26257/flipr_test>.

Sticking with simply C<db:pg:flipr_test> for these examples, we use the

C<deploy> command to deploy the change:

> sqitch deploy db:pg:flipr_test

Adding registry tables to db:pg:flipr_test

Deploying to db:pg:flipr_test

+ appschema .. ok

First Sqitch created registry tables used to track database changes. The

structure and name of the registry varies between databases (PostgreSQL uses a

schema to namespace its registry, while SQLite and MySQL use separate

databases). Next, Sqitch deploys changes. We only have one so far; the C<+>

reinforces the idea that the change is being C<added> to the database.

With this change deployed, if you connect to the database, you'll be able to

see the schema:

> psql -d flipr_test -c '\dn flipr'

List of schemas

Name | Owner

-------+-------

flipr | marge

=head2 Trust, But Verify

But that's too much work. Do you really want to do something like that after

every deploy?

Here's where the C<verify> script comes in. Its job is to test that the deploy

did what it was supposed to. It should do so without regard to any data that

might be in the database, and should throw an error if the deploy was not

successful. In PostgreSQL, the simplest way to do so for non-queryable objects

such as schemas is to take advantage the

L<access privilege inquiry functions|https://www.postgresql.org/docs/current/static/functions-info.html#FUNCTIONS-INFO-ACCESS-TABLE>.

These functions conveniently throw exceptions if the object being inquired

does not exist. For our new schema, C<has_schema_privilege()> will do very

nicely. Put this query into F<verify/appschema.sql>:

SELECT pg_catalog.has_schema_privilege('flipr', 'usage');

B<Important!> This query isn't verifying that the user has C<usage> privilege

on schema C<flipr>. The verification will pass even if the current user

has no usage rights.

B<Important!> Both C<SELECT false;> and C<SELECT true;> queries will successfully

pass C<verify> step. Only queries that raise an exception will fail.

Such functionality may not be available to other databases, but you can use

I<any> query that will throw an exception if the schema doesn't exist. One

handy way to do that is to divide by zero if an object doesn't exist. So for

other databases, assuming division by zero is fatal, you could do something

like this:

SELECT 1/COUNT(*) FROM information_schema.schemata WHERE schema_name = 'flipr';

In Postgres 9.5+ you can use C<PL/pgSQL> anonymous functions with

C<ASSERT> / C<RAISE> statements.

DO $$

BEGIN

ASSERT (SELECT has_schema_privilege('flipr', 'usage'));

END $$;

You can use variables to perform more complex checks:

DO $$

DECLARE

result varchar;

BEGIN

result := (SELECT name FROM flipr.pipelines WHERE id = 1);

ASSERT result = 'Example';

END $$;

This example ensures the record with C<id=1> in C<pipelines> table

has C<name> field equals C<'Example'>.

Either way, run the C<verify> script with the L<C<verify>|sqitch-verify>

command:

> sqitch verify db:pg:flipr_test

Verifying db:pg:flipr_test

* appschema .. ok

Verify successful

Looks good! If you want to make sure that the verify script correctly dies if

the schema doesn't exist, temporarily change the schema name in the script to

something that doesn't exist, something like:

SELECT pg_catalog.has_schema_privilege('nonesuch', 'usage');

Then L<C<verify>|sqitch-verify> again:

> sqitch verify db:pg:flipr_test

Verifying db:pg:flipr_test

* appschema .. psql:verify/appschema.sql:5: ERROR: schema "nonesuch" does not exist

# Verify script "verify/appschema.sql" failed.

not ok

Verify Summary Report

---------------------

Changes: 1

Errors: 1

Verify failed

It's even nice enough to tell us what the problem is. Or, for the

divide-by-zero example, change the schema name:

SELECT 1/COUNT(*) FROM information_schema.schemata WHERE schema_name = 'nonesuch';

Then the verify will look something like:

> sqitch verify db:pg:flipr_test

Verifying db:pg:flipr_test

* appschema .. psql:verify/appschema.sql:5: ERROR: division by zero

# Verify script "verify/appschema.sql" failed.

not ok

Verify Summary Report

---------------------

Changes: 1

Errors: 1

Verify failed

Less useful error output, but enough to alert us that something has gone

wrong.

Don't forget to change the schema name back before continuing!

=head2 Status, Revert, Log, Repeat

For purely informational purposes, we can always see how a deployment was

recorded via the L<C<status>|sqitch-status> command, which reads the registry

tables from the database:

> sqitch status db:pg:flipr_test

# On database db:pg:flipr_test

# Project: flipr

# Change: c7981df861183412b01be706889e508a63d445ca

# Name: appschema

# Deployed: 2013-12-30 15:27:15 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Let's make sure that we can revert the change:

> sqitch revert db:pg:flipr_test

Revert all changes from db:pg:flipr_test? [Yes]

- appschema .. ok

The L<C<revert>|sqitch-revert> command first prompts to make sure that we

really do want to revert. This is to prevent unnecessary accidents. You can

pass the C<-y> option to disable the prompt. Also, notice the C<-> before the

change name in the output, which reinforces that the change is being

I<removed> from the database. And now the schema should be gone:

> psql -d flipr_test -c '\dn flipr'

List of schemas

Name | Owner

------+-------

And the status message should reflect as much:

> sqitch status db:pg:flipr_test

# On database db:pg:flipr_test

No changes deployed

Of course, since nothing is deployed, the L<C<verify>|sqitch-verify> command

has nothing to verify:

> sqitch verify db:pg:flipr_test

Verifying db:pg:flipr_test

No changes deployed

However, we still have a record that the change happened, visible via the

L<C<log>|sqitch-log> command:

> sqitch log db:pg:flipr_test

On database db:pg:flipr_test

Revert c7981df861183412b01be706889e508a63d445ca

Name: appschema

Committer: Marge N. O’Vera <marge@example.com>

Date: 2013-12-30 15:38:17 -0800

Add schema for all flipr objects.

Deploy c7981df861183412b01be706889e508a63d445ca

Name: appschema

Committer: Marge N. O’Vera <marge@example.com>

Date: 2013-12-30 15:27:15 -0800

Add schema for all flipr objects.

Note that the actions we took are shown in reverse chronological order, with

the revert first and then the deploy.

Cool. Now let's commit it.

> git add .

> git commit -m 'Add flipr schema.'

[main d812132] Add flipr schema.

4 files changed, 22 insertions(+)

create mode 100644 deploy/appschema.sql

create mode 100644 revert/appschema.sql

create mode 100644 verify/appschema.sql

And then deploy again. This time, let's use the C<--verify> option, so that

the C<verify> script is applied when the change is deployed:

> sqitch deploy --verify db:pg:flipr_test

Deploying changes to db:pg:flipr_test

+ appschema .. ok

And now the schema should be back:

> psql -d flipr_test -c '\dn flipr'

List of schemas

Name | Owner

-------+-------

flipr | marge

When we look at the status, the deployment will be there:

> sqitch status db:pg:flipr_test

# On database db:pg:flipr_test

# Project: flipr

# Change: c7981df861183412b01be706889e508a63d445ca

# Name: appschema

# Deployed: 2013-12-30 15:40:53 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

=head1 On Target

I'm getting a little tired of always having to type C<db:pg:flipr_test>,

aren't you? This L<database connection URI|https://github.com/libwww-perl/uri-db/>

tells Sqitch how to connect to the deployment target, but we don't have

to keep using the URI. We can name the target:

> sqitch target add flipr_test db:pg:flipr_test

The L<C<target>|sqitch-target> command, inspired by

L<C<git-remote>|https://git-scm.com/docs/git-remote>, allows management of one

or more named deployment targets. We've just added a target named

C<flipr_test>, which means we can use the string C<flipr_test> for the target,

rather than the URI. But since we're doing so much testing, we can also use

the L<C<engine>|sqitch-engine> command to tell Sqitch to deploy to the

C<flipr_test> target by default:

> sqitch engine add pg flipr_test

Now we can omit the target argument altogether, unless we need to deploy to

another database. Which we will, eventually, but at least our examples will be

simpler from here on in, e.g.:

> sqitch status

# On database flipr_test

# Project: flipr

# Change: c7981df861183412b01be706889e508a63d445ca

# Name: appschema

# Deployed: 2013-12-30 15:40:53 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Yay, that allows things to be a little more concise. Let's also make sure that

changes are verified after deploying them:

> sqitch config --bool deploy.verify true

> sqitch config --bool rebase.verify true

We'll see the L<C<rebase>|sqitch-rebase> command a bit later. In the meantime,

let's commit the new configuration and and make some more changes!

> git commit -am 'Set default deployment target and always verify.'

[main a6267d3] Set default deployment target and always verify.

1 file changed, 8 insertions(+)

=head1 Deploy with Dependency

Let's add another change, this time to create a table. Our app will need

users, of course, so we'll create a table for them. First, add the new change:

> sqitch add users --requires appschema -n 'Creates table to track our users.'

Created deploy/users.sql

Created revert/users.sql

Created verify/users.sql

Added "users [appschema]" to sqitch.plan

Note that we're requiring the C<appschema> change as a dependency of the new

C<users> change. Although that change has already been added to the plan and

therefore should always be applied before the C<users> change, it's a good

idea to be explicit about dependencies.

Now edit the scripts. When you're done, F<deploy/users.sql> should look like

this:

-- Deploy flipr:users to pg

-- requires: appschema

BEGIN;

SET client_min_messages = 'warning';

CREATE TABLE flipr.users (

nickname TEXT PRIMARY KEY,

password TEXT NOT NULL,

timestamp TIMESTAMPTZ NOT NULL DEFAULT NOW()

);

COMMIT;

A few things to notice here. On the second line, the dependence on the

C<appschema> change has been listed. This doesn't do anything, but the default

C<deploy> PostgreSQL template lists it here for your reference while editing

the file. Useful, right?

Notice that all of the SQL code is wrapped in a transaction. This is handy for

PostgreSQL deployments, because PostgreSQL DDLs are transactional. The upshot

is that if any part of this deploy script fails, the whole change fails. Such

may work less-well for database engines that don't support transactional DDLs.

The table itself will be created in the C<flipr> schema. This is why we need

to require the C<appschema> change.

Now for the verify script. The simplest way to check that the table was

created and has the expected columns without touching the data? Just select

from the table with a false C<WHERE> clause. Add this to F<verify/users.sql>:

SELECT nickname, password, timestamp

FROM flipr.users

WHERE FALSE;

Now for the revert script: all we have to do is drop the table. Add this to

F<revert/users.sql>:

DROP TABLE flipr.users;

Couldn't be much simpler, right? Let's deploy this bad boy:

> sqitch deploy

Deploying changes to flipr_test

+ users .. ok

We know, since verification is enabled, that the table must have been created.

But for the purposes of visibility, let's have a quick look:

> psql -d flipr_test -c '\d flipr.users'

Table "flipr.users"

Column | Type | Modifiers

-----------+--------------------------+------------------------

nickname | text | not null

password | text | not null

timestamp | timestamp with time zone | not null default now()

Indexes:

"users_pkey" PRIMARY KEY, btree (nickname)

We can also verify all currently deployed changes with the

L<C<verify>|sqitch-verify> command:

> sqitch verify

Verifying flipr_test

* appschema .. ok

* users ...... ok

Verify successful

Now have a look at the status:

> sqitch status

# On database flipr_test

# Project: flipr

# Change: 77398e1dbc5fbce58b05eb67d201f15774718727

# Name: users

# Deployed: 2013-12-30 15:51:09 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Success! Let's make sure we can revert the change, as well:

> sqitch revert --to @HEAD^ -y

Reverting changes to appschema from flipr_test

- users .. ok

Note that we've used the C<--to> option to specify the change to revert to.

And what do we revert to? The symbolic tag C<@HEAD>, when passed to

L<C<revert>|sqitch-revert>, always refers to the last change deployed to the

database. (For other commands, it refers to the last change in the plan.)

Appending the caret (C<^>) tells Sqitch to select the change I<prior> to the

last deployed change. So we revert to C<appschema>, the penultimate change.

The other potentially useful symbolic tag is C<@ROOT>, which refers to the

first change deployed to the database (or in the plan, depending on the

command).

Back to the database. The C<users> table should be gone but the C<flipr> schema

should still be around:

> psql -d flipr_test -c '\d flipr.users'

Did not find any relation named "flipr.users".

The L<C<status>|sqitch-status> command politely informs us that we have

undeployed changes:

> sqitch status

# On database flipr_test

# Project: flipr

# Change: c7981df861183412b01be706889e508a63d445ca

# Name: appschema

# Deployed: 2013-12-30 15:40:53 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Undeployed change:

* users

As does the L<C<verify>|sqitch-verify> command:

> sqitch verify

Verifying flipr_test

* appschema .. ok

Undeployed change:

* users

Verify successful

Note that the verify is successful, because all currently-deployed changes are

verified. The list of undeployed changes (just "users" here) reminds us about

the current state.

Okay, let's commit and deploy again:

> git add .

> git commit -am 'Add users table.'

[main d58ea2f] Add users table.

4 files changed, 31 insertions(+)

create mode 100644 deploy/users.sql

create mode 100644 revert/users.sql

create mode 100644 verify/users.sql

> sqitch deploy

Deploying changes to flipr_test

+ users .. ok

Looks good. Check the status:

> sqitch status

# On database flipr_test

# Project: flipr

# Change: 77398e1dbc5fbce58b05eb67d201f15774718727

# Name: users

# Deployed: 2013-12-30 15:57:14 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Excellent. Let's do some more!

=head1 Add Two at Once

Let's add a couple more changes to add functions for managing users.

> sqitch add insert_user --requires users --requires appschema \

-n 'Creates a function to insert a user.'

Created deploy/insert_user.sql

Created revert/insert_user.sql

Created verify/insert_user.sql

Added "insert_user [users appschema]" to sqitch.plan

> sqitch add change_pass --requires users --requires appschema \

-n 'Creates a function to change a user password.'

Created deploy/change_pass.sql

Created revert/change_pass.sql

Created verify/change_pass.sql

Added "change_pass [users appschema]" to sqitch.plan

Now might be a good time to have a look at the deployment plan:

> cat sqitch.plan

%syntax-version=1.0.0

%project=flipr

%uri=https://github.com/sqitchers/sqitch-intro/

appschema 2013-12-30T23:19:45Z Marge N. O’Vera <marge@example.com> # Add schema for all flipr objects.

users [appschema] 2013-12-30T23:49:00Z Marge N. O’Vera <marge@example.com> # Creates table to track our users.

insert_user [users appschema] 2013-12-30T23:57:36Z Marge N. O’Vera <marge@example.com> # Creates a function to insert a user.

change_pass [users appschema] 2013-12-30T23:57:45Z Marge N. O’Vera <marge@example.com> # Creates a function to change a user password.

Each change appears on a single line with the name of the change, a bracketed

list of dependencies, a timestamp, the name and email address of the user who

planned the change, and a note.

Let's write the code for the new changes. Here's what

F<deploy/insert_user.sql> should look like:

-- Deploy flipr:insert_user to pg

-- requires: users

-- requires: appschema

BEGIN;

CREATE OR REPLACE FUNCTION flipr.insert_user(

nickname TEXT,

password TEXT

) RETURNS VOID LANGUAGE SQL SECURITY DEFINER AS $$

INSERT INTO flipr.users VALUES($1, md5($2));

$$;

COMMIT;

Here's what F<verify/insert_user.sql> might look like:

BEGIN;

SELECT has_function_privilege('flipr.insert_user(text, text)', 'execute');

ROLLBACK;

We simply take advantage of the fact that C<has_function_privilege()> throws

an exception if the specified function does not exist.

And F<revert/insert_user.sql> should look something like this:

-- Revert flipr:insert_user from pg

BEGIN;

DROP FUNCTION flipr.insert_user(TEXT, TEXT);

COMMIT;

Now for C<change_pass>; F<deploy/change_pass.sql> might look like this:

-- Deploy flipr:change_pass to pg

-- requires: users

-- requires: appschema

BEGIN;

CREATE OR REPLACE FUNCTION flipr.change_pass(

nick TEXT,

oldpass TEXT,

newpass TEXT

) RETURNS BOOLEAN LANGUAGE plpgsql SECURITY DEFINER AS $$

BEGIN

UPDATE flipr.users

SET password = md5($3)

WHERE nickname = $1

AND password = md5($2);

RETURN FOUND;

END;

$$;

COMMIT;

Use C<has_function_privilege()> in F<verify/change_pass.sql> again:

BEGIN;

SELECT has_function_privilege('flipr.change_pass(text, text, text)', 'execute');

ROLLBACK;

And of course, its C<revert> script, F<revert/change_pass.sql>, should look

something like:

-- Revert flipr:change_pass from pg

BEGIN;

DROP FUNCTION flipr.change_pass(TEXT, TEXT, TEXT);

COMMIT;

Try em out!

> sqitch deploy

Deploying changes to flipr_test

+ insert_user .. ok

+ change_pass .. ok

Do we have the functions? Of course we do, they were verified. Still, have a

look:

> psql -d flipr_test -c '\df flipr.*'

List of functions

Schema | Name | Result data type | Argument data types | Type

--------+-------------+------------------+---------------------------------------+--------

flipr | change_pass | boolean | nick text, oldpass text, newpass text | normal

flipr | insert_user | void | nickname text, password text | normal

And what's the status?

> sqitch status

# On database flipr_test

# Project: flipr

# Change: 01a4f6964b89284525cb5877d222df8be70d1647

# Name: change_pass

# Deployed: 2013-12-30 15:59:44 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Looks good. Let's make sure revert works:

> sqitch revert -y --to @HEAD^^

Reverting changes to users from flipr_test

- change_pass .. ok

- insert_user .. ok

> psql -d flipr_test -c '\df flipr.*'

List of functions

Schema | Name | Result data type | Argument data types | Type

--------+------+------------------+---------------------+------

Note the use of C<@HEAD^^> to specify that the revert be to two changes prior

the last deployed change. Looks good. Let's do the commit and re-deploy dance:

> git add .

> git commit -m 'Add `insert_user()` and `change_pass()`.'

[main c9b4d68] Add `insert_user()` and `change_pass()`.

7 files changed, 65 insertions(+)

create mode 100644 deploy/change_pass.sql

create mode 100644 deploy/insert_user.sql

create mode 100644 revert/change_pass.sql

create mode 100644 revert/insert_user.sql

create mode 100644 verify/change_pass.sql

create mode 100644 verify/insert_user.sql

> sqitch deploy

Deploying changes to flipr_test

+ insert_user .. ok

+ change_pass .. ok

> sqitch status

# On database flipr_test

# Project: flipr

# Change: 01a4f6964b89284525cb5877d222df8be70d1647

# Name: change_pass

# Deployed: 2013-12-30 16:00:50 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

> sqitch verify

Verifying flipr_test

* appschema .... ok

* users ........ ok

* insert_user .. ok

* change_pass .. ok

Verify successful

Great, we're fully up-to-date!

=head1 Ship It!

Let's do a first release of our app. Let's call it C<1.0.0-dev1> Since we want

to have it go out with deployments tied to the release, let's tag it:

> sqitch tag v1.0.0-dev1 -n 'Tag v1.0.0-dev1.'

Tagged "change_pass" with @v1.0.0-dev1

> git commit -am 'Tag the database with v1.0.0-dev1.'

[main 0acef3e] Tag the database with v1.0.0-dev1.

1 file changed, 1 insertion(+)

> git tag v1.0.0-dev1 -am 'Tag v1.0.0-dev1'

We can try deploying to make sure the tag gets picked up like so:

> createdb flipr_dev

> sqitch deploy db:pg:flipr_dev

Adding registry tables to db:pg:flipr_dev

Deploying changes to db:pg:flipr_dev

+ appschema ................. ok

+ users ..................... ok

+ insert_user ............... ok

+ change_pass @v1.0.0-dev1 .. ok

Great, all four changes were deployed and C<change_pass> was tagged with

C<@v1.0.0-dev1>. Let's have a look at the status:

> sqitch status db:pg:flipr_dev

# On database db:pg:flipr_dev

# Project: flipr

# Change: 01a4f6964b89284525cb5877d222df8be70d1647

# Name: change_pass

# Tag: @v1.0.0-dev1

# Deployed: 2013-12-30 16:02:19 -0800

# By: Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Note the listing of the tag as part of the status message. Now let's bundle

everything up for release:

> sqitch bundle

Bundling into bundle/

Writing config

Writing plan

Writing scripts

+ appschema

+ users

+ insert_user

+ change_pass @v1.0.0-dev1

Now we can package the F<bundle> directory and distribute it. When it gets

installed somewhere, users can use Sqitch to deploy to the database. Let's try

deploying it:

> cd bundle

> createdb flipr_prod

> sqitch deploy db:pg:flipr_prod

Adding registry tables to db:pg:flipr_prod

Deploying changes to db:pg:flipr_prod

+ appschema ................. ok

+ users ..................... ok

+ insert_user ............... ok

+ change_pass @v1.0.0-dev1 .. ok

Looks much the same as before, eh? Package it up and ship it!

> cd ..

> mv bundle flipr-v1.0.0-dev1

> tar -czf flipr-v1.0.0-dev1.tgz flipr-v1.0.0-dev1

=head1 Flip Out

Now that we've got the basics of user management done, let's get to work on

the core of our product, the "flip." Since other folks are working on other

tasks in the repository, we'll work on a branch, so we can all stay out of

each other's way. So let's branch:

> git checkout -b flips

Switched to a new branch 'flips'

Now we can add a new change to create a table for our flips.

> sqitch add flips -r appschema -r users -n 'Adds table for storing flips.'

Created deploy/flips.sql

Created revert/flips.sql

Created verify/flips.sql

Added "flips [appschema users]" to sqitch.plan

You know the drill by now. Edit F<deploy/flips.sql>:

-- Deploy flipr:flips to pg

-- requires: appschema

-- requires: users

BEGIN;

SET client_min_messages = 'warning';

CREATE TABLE flipr.flips (

id BIGSERIAL PRIMARY KEY,

nickname TEXT NOT NULL REFERENCES flipr.users(nickname),

body TEXT NOT NULL DEFAULT '' CHECK ( length(body) <= 180 ),

timestamp TIMESTAMPTZ NOT NULL DEFAULT clock_timestamp()

);

COMMIT;

Edit F<verify/flips.sql>:

-- Verify flipr:flips on pg

BEGIN;

SELECT id

, nickname

, body

, timestamp

FROM flipr.flips

WHERE FALSE;

ROLLBACK;

And edit F<revert/flips.sql>:

-- Revert flipr:flips from pg

BEGIN;

DROP TABLE flipr.flips;

COMMIT;

And give it a whirl:

> sqitch deploy

Deploying changes to flipr_test

+ flips .. ok

Look good?

> sqitch status --show-tags

# On database flipr_test

# Project: flipr

# Change: 4d164ef5986450f00a565735518b1d126f8ee69d

# Name: flips

# Deployed: 2013-12-30 16:34:38 -0800

# By: Marge N. O’Vera <marge@example.com>

#

# Tag:

# @v1.0.0-dev1 - 2013-12-30 16:34:38 -0800 - Marge N. O’Vera <marge@example.com>

#

Nothing to deploy (up-to-date)

Note the use of C<--show-tags> to show all the deployed tags. Now make it so:

> git add .

[flips e8f4655] Add flips table.

> git commit -am 'Add flips table.'

4 files changed, 37 insertions(+)

create mode 100644 deploy/flips.sql

create mode 100644 revert/flips.sql

create mode 100644 verify/flips.sql

=head1 Wash, Rinse, Repeat

Now comes the time to add functions to manage flips. I'm sure you have things

nailed down now. Go ahead and add C<insert_flip> and C<delete_flip> changes

and commit them. The C<insert_flip> deploy script might look something like:

-- Deploy flipr:insert_flip to pg

-- requires: flips

-- requires: appschema

-- requires: users

BEGIN;

CREATE OR REPLACE FUNCTION flipr.insert_flip(

nickname TEXT,

body TEXT

) RETURNS BIGINT LANGUAGE sql SECURITY DEFINER AS $$

INSERT INTO flipr.flips (nickname, body)

VALUES ($1, $2)

RETURNING id;