Provided by: sqitch_1.3.1-1_all 
Name
sqitchtutorial-exasol - A tutorial introduction to Sqitch change management on Exasol
Synopsis
sqitch *
Description
This tutorial explains how to create a sqitch-enabled Exasol 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 Git <https://git-scm.com/> as the VCS and Exasol
<https://www.exasol.com/> as the storage engine, but for the most part you can substitute
other VCSes and database engines in the examples as appropriate.
If you'd like to manage a PostgreSQL database, see sqitchtutorial.
If you'd like to manage an SQLite database, see sqitchtutorial-sqlite.
If you'd like to manage an Oracle database, see sqitchtutorial-oracle.
If you'd like to manage a MySQL database, see sqitchtutorial-mysql.
If you'd like to manage a Firebird database, see sqitchtutorial-firebird.
If you'd like to manage a Vertica database, see sqitchtutorial-vertica.
If you'd like to manage a Snowflake database, see sqitchtutorial-snowflake.
Connection Configuration
Sqitch requires ODBC to connect to the Exasol database. As such, you'll need to make sure
that the Exasol ODBC driver is properly configured. At its simplest, on Unix-like systems,
name the driver "Exasol" by adding this entry to "odbcinst.ini" (usually found in "/etc",
"/usr/etc", or "/usr/local/etc"):
[Exasol]
Description = ODBC for Exasol
Driver = /opt/EXASOL_ODBC-6.0.4/lib/linux/x86_64/libexaodbc-uo2214lv2.so
Note that you'll need to adjust the path depending on the version of the ODBC driver, and
where you installed it.
You might also consider naming your database connection by putting an entry in
"~/.odbc.ini", like so (assuming that Exasol is running on your local host):
[flipr_test]
Driver = Exasol
EXAHOST = 127.0.0.1:8563
EXAUID = sys
EXAPWD = exasol
Putting user and password information here is optional, but probably safer than other
available options as long as the file is protected (mode 0600) so that only you can read
it.
Normally, of course, you'd have a separate user per project (and a separate ODBC
connection defined); the above example is taken from the simplified tutorial setup we're
using.
See the Exasol downloads <https://www.exasol.com/portal/display/DOWNLOAD/6.0> for ODBC
drivers, documentation, etc.
Database Setup
While installing and configuring an Exasol instance is beyond the scope of this tutorial,
if you just want to follow along, you can start a Docker instance:
> docker run --detach --privileged --stop-timeout 120 -p 127.0.0.1:8563:8888 exasol/docker-db:latest
This will make Exasol available at 127.0.0.1 (localhost) on port 8563, as if you were
running it directly on your machine. You may have to wait for a minute until the database
finished initializing; you can try connecting with `exaplus` to check status.
See the corresponding GitHub repository <https://github.com/EXASOL/docker-db> for more
information about how to run Exasol using Docker, for example if you want your data to
persist beyond the lifetime of the container, etc.
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 on GitHub <https://github.com/sqitchers/sqitch-exasol-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-exasol-intro/ --engine exasol
Created sqitch.conf
Created sqitch.plan
Created deploy/
Created revert/
Created verify/
Let's have a look at sqitch.conf:
> cat sqitch.conf
[core]
engine = exasol
# plan_file = sqitch.plan
# top_dir = .
# [engine "exasol"]
# target = db:exasol:
# registry = sqitch
# client = exaplus
Good, it picked up on the fact that we're creating changes for the Exasol engine, thanks
to the "--engine exasol" option, and saved it to the file. Furthermore, it wrote a
commented-out "[engine "exasol"]" section with all the available Exasol engine-specific
settings commented out and ready to be edited as appropriate.
By default, Sqitch will read sqitch.conf in the current directory for settings. But it
will also read ~/.sqitch/sqitch.conf for user-specific settings. Since Exasol's "exaplus"
client is not in the path on my system, let's go ahead an tell it where to find the client
on our computer (don't bother if you're using the 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.exasol.client /opt/EXAplus/exaplus
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. OXVera'
> sqitch config --user user.email 'marge@example.com'
Have a look at ~/.sqitch/sqitch.conf and you'll see this:
> cat ~/.sqitch/sqitch.conf
[engine "exasol"]
client = /opt/EXAplus/exaplus
[user]
name = Marge N. OXVera
email = marge@example.com
Which means that Sqitch should be able to find "exaplus" 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, sqitch.plan:
> cat sqitch.plan
%syntax-version=1.0.0
%project=flipr
%uri=https://github.com/sqitchers/sqitch-exasol-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 "%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 a42564d] Initialize Sqitch configuration.
2 files changed, 16 insertions(+), 0 deletions(-)
create mode 100644 sqitch.conf
create mode 100644 sqitch.plan
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 "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 "deploy" script's job is
to create the schema. So we add this to deploy/appschema.sql:
CREATE SCHEMA flipr;
The "revert" script's job is to precisely revert the change to the deploy script, so we
add this to revert/appschema.sql:
DROP SCHEMA flipr;
Now we can try deploying this change. We tell Sqitch where to send the change via a
database URI <https://github.com/libwww-perl/uri-db/>, assuming the default "sys" user and
an ODBC driver named "Exasol" (see "Connection Configuration" for details):
> sqitch deploy 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Adding registry tables to db:exasol://sys:@localhost:8563/?Driver=Exasol
Deploying changes to db:exasol://sys:@localhost:8563/?Driver=Exasol
+ appschema .. ok
(If you see an error resolving the TLS certificate, you can disable certificate
verification by adding "&SSLCERTIFICATE=SSL_VERIFY_NONE" to the end of the URL. Only do
this for testing!)
First Sqitch created registry tables used to track database changes. The structure and
name of the registry varies between databases (Exasol 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 "+" reinforces the idea that the change is being "added" to the
database.
With this change deployed, if you connect to the database, you'll be able to see the
schema:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "select schema_name from exa_schemas;"
SCHEMA_NAME
--------------------------------------------------------------------------------------------------------------------------------
SQITCH
FLIPR
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 "verify" script comes in. Its job is to test that the deploy did was 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 Exasol, the simplest way to
do so for schema is probably to simply create an object in the schema. Put this SQL into
verify/appschema.sql:
CREATE TABLE flipr.verify__ (id int);
DROP TABLE flipr.verify__;
In truth, you can use any query that generates an SQL error if the schema doesn't exist.
Another handy way to do that is to divide by zero if an object doesn't exist. For example,
to throw an error when the "flipr" schema does not exist, you could do something like
this:
SELECT 1/COUNT(*) FROM exa_schemas WHERE schema_name = 'FLIPR';
Either way, run the "verify" script with the "verify" command:
> sqitch verify 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Verifying db:exasol://sys:@localhost:8563/?Driver=Exasol
* 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:
CREATE TABLE nonesuch.verify__ (id int);
Then "verify" again:
> sqitch verify 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Verifying db:exasol://sys:@localhost:8563/?Driver=Exasol
* appschema .. Error: [42000] schema NONESUCH not found [line 1, column 40] (Session: 1582884049218108749)
# 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 exa_schemas WHERE schema_name = 'nonesuch';
Then the verify will look something like:
> sqitch verify 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Verifying db:exasol://sys:@localhost:8563/?Driver=Exasol
* appschema .. Error: [22012] data exception - division by zero (Session: 1582884446489810101)
# 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!
Status, Revert, Log, Repeat
For purely informational purposes, we can always see how a deployment was recorded via the
"status" command, which reads the registry tables from the database:
> sqitch status 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
# On database db:exasol://sys:@localhost:8563/?Driver=Exasol
# Project: flipr
# Change: f9759f0ed77964b6a3b6c7aa3b6058b4bb7db764
# Name: appschema
# Deployed: 2014-09-04 15:26:28 -0700
# By: Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
Let's make sure that we can revert the change:
> sqitch revert 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Revert all changes from db:exasol://sys:@localhost:8563/?Driver=Exasol? [Yes]
- appschema .. ok
The "revert" command first prompts to make sure that we really do want to revert. This is
to prevent unnecessary accidents. You can pass the "-y" option to disable the prompt.
Also, notice the "-" before the change name in the output, which reinforces that the
change is being removed from the database. And now the schema should be gone:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "select schema_name from
exa_schemas;"
SCHEMA_NAME
--------------------------------------------------------------------------------------------------------------------------------
SQITCH
And the status message should reflect as much:
> sqitch status 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
# On database db:exasol://sys:@localhost:8563/?Driver=Exasol
No changes deployed
Of course, since nothing is deployed, the "verify" command has nothing to verify:
> sqitch verify 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Verifying db:exasol://sys:@localhost:8563/?Driver=Exasol
No changes deployed
However, we still have a record that the change happened, visible via the "log" command:
> sqitch log 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
On database db:exasol://sys:@localhost:8563/?Driver=Exasol
Revert f9759f0ed77964b6a3b6c7aa3b6058b4bb7db764
Name: appschema
Committer: Marge N. OXVera <marge@example.com>
Date: 2014-09-04 16:33:02 -0700
Add schema for all flipr objects.
Deploy f9759f0ed77964b6a3b6c7aa3b6058b4bb7db764
Name: appschema
Committer: Marge N. OXVera <marge@example.com>
Date: 2014-09-04 15:26:28 -0700
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 9bee4bd] Add flipr schema.
5 files changed, 197 insertions(+), 0 deletions(-)
create mode 100644 deploy/appschema.sql
create mode 100644 revert/appschema.sql
create mode 100644 sqitch.sql
create mode 100644 verify/appschema.sql
And then deploy again. This time, let's use the "--verify" option, so that the "verify"
script is applied when the change is deployed:
> sqitch deploy --verify 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
Deploying changes to db:exasol://sys:@localhost:8563/?Driver=Exasol
+ appschema .. ok
And now the schema should be back:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "select schema_name from exa_schemas;"
SCHEMA_NAME
--------------------------------------------------------------------------------------------------------------------------------
SQITCH
FLIPR
When we look at the status, the deployment will be there:
> sqitch status 'db:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
# On database db:exasol://sys:@localhost:8563/?Driver=Exasol
# Project: flipr
# Change: fef4c2911ae68aee8f6ea164293a32923dc13b67
# Name: appschema
# Deployed: 2014-09-04 16:37:38 -0700
# By: Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
On Target
I'm getting a little tired of always having to type
"db:exasol://sys:exasol@localhost:8563/?Driver=Exasol", aren't you? This 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:exasol://sys:exasol@localhost:8563/?Driver=Exasol'
The "target" command, inspired by "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
"flipr_test", which means we can use the string "flipr_test" for the target, rather than
the URI. But since we're doing so much testing, we can also tell Sqitch to deploy to the
"flipr_test" target by default:
> sqitch engine add exasol 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: f9759f0ed77964b6a3b6c7aa3b6058b4bb7db764
# Name: appschema
# Deployed: 2014-09-04 16:37:38 -0700
# By: Marge N. OXVera <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 "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 469779a] Set default deployment target and always verify.
1 files changed, 8 insertions(+), 0 deletions(-)
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 "appschema" change as a dependency of the new "users"
change. Although that change has already been added to the plan and therefore should
always be applied before the "users" change, it's a good idea to be explicit about
dependencies.
Now edit the scripts. When you're done, deploy/users.sql should look like this:
-- Deploy flipr:users to exasol
-- requires: appschema
CREATE TABLE flipr.users (
nickname VARCHAR(64) PRIMARY KEY,
password VARCHAR(256) NOT NULL,
fullname VARCHAR(256) NOT NULL,
twitter VARCHAR(256) NOT NULL,
ts TIMESTAMP WITH LOCAL TIME ZONE DEFAULT NOW() NOT NULL
);
COMMIT;
A few things to notice here. On the second line, the dependence on the "appschema" change
has been listed. This doesn't do anything, but the default "deploy" Exasol template lists
it here for your reference while editing the file. Useful, right?
The table itself will be created in the "flipr" schema. This is why we need to require the
"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
"WHERE" clause. Add this to verify/users.sql:
SELECT nickname, password, fullname, twitter, ts
FROM flipr.users
WHERE FALSE;
Now for the revert script: all we have to do is drop the table. Add this to
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:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.users;"
COLUMN_NAME SQL_TYPE NULLABLE DISTRIBUTION_KEY
-------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- -------- ----------------
NICKNAME VARCHAR(64) UTF8 FALSE FALSE
PASSWORD VARCHAR(256) UTF8 FALSE FALSE
FULLNAME VARCHAR(256) UTF8 FALSE FALSE
TWITTER VARCHAR(256) UTF8 FALSE FALSE
TS TIMESTAMP WITH LOCAL TIME ZONE FALSE FALSE
We can also verify all currently deployed changes with the "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: 794a6c78816543909d592e2e9f5c0fade5b47406
# Name: users
# Deployed: 2017-11-02 11:02:40 +0100
# By: Marge N. OXVera <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 "--to" option to specify the change to revert to. And what do we
revert to? The symbolic tag @HEAD, when passed to "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 ("^") tells Sqitch to select the change prior to the last
deployed change. So we revert to "appschema", the penultimate change. The other
potentially useful symbolic tag is @ROOT, which refers to the first change deployed to the
database (or in the plan, depending on the command).
Back to the database. The "users" table should be gone but the "flipr" schema should still
be around:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.users;"
Error: [42000] table or view FLIPR.USERS not found [line 1, column 10] (Session: 1582958508294847446)
The "status" command politely informs us that we have undeployed changes:
> sqitch status
# On database flipr_test
# Project: flipr
# Change: f9759f0ed77964b6a3b6c7aa3b6058b4bb7db764
# Name: appschema
# Deployed: 2014-09-04 16:37:38 -0700
# By: Marge N. OXVera <marge@example.com>
#
Undeployed change:
* users
As does the "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 c7c24c5] Add users table.
4 files changed, 18 insertions(+), 0 deletions(-)
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: d647ac8c130a7e0b12c9049789e46afb4a4f6e53
# Name: users
# Deployed: 2014-09-04 17:42:53 -0700
# By: Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
Excellent. Let's do some more!
Add Two at Once
Let's add a couple more changes. Our app will need to store status messages from users.
Let's call them -- and the table to store them -- "flips". And we'll also need a view that
lists user names with their flips. Let's add changes for them both:
> 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
> sqitch add userflips -r appschema -r users -r flips \
-n 'Creates the userflips view.'
Created deploy/userflips.sql
Created revert/userflips.sql
Created verify/userflips.sql
Added "userflips [appschema users flips]" 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-exasol-intro/
appschema 2014-09-04T18:40:34Z Marge N. OXVera <marge@example.com> # Add schema for all flipr objects.
users [appschema] 2014-09-04T23:40:15Z Marge N. OXVera <marge@example.com> # Creates table to track our users.
flips [appschema users] 2014-09-05T00:16:58Z Marge N. OXVera <marge@example.com> # Adds table for storing flips.
userflips [appschema users flips] 2014-09-05T00:18:43Z Marge N. OXVera <marge@example.com> # Creates the userflips view.
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 deploy/flips.sql should look like:
-- Deploy flipr:flips to exasol
-- requires: appschema
-- requires: users
CREATE TABLE flipr.flips (
id INTEGER IDENTITY PRIMARY KEY,
nickname VARCHAR(64) NOT NULL REFERENCES flipr.users(nickname),
body VARCHAR(180) DEFAULT '' NOT NULL,
ts TIMESTAMP WITH LOCAL TIME ZONE DEFAULT NOW() NOT NULL
);
COMMIT;
Here's what verify/flips.sql might look like:
-- Verify flipr:flips on exasol
SELECT id, nickname, body, ts
FROM flipr.flips
WHERE FALSE;
ROLLBACK;
And revert/flips.sql should look something like this:
-- Revert flipr:flips from exasol
DROP TABLE flipr.flips;
COMMIT;
Now for "userflips"; deploy/userflips.sql might look like this:
-- Deploy flipr:userflips to exasol
-- requires: appschema
-- requires: users
-- requires: flips
CREATE OR REPLACE VIEW flipr.userflips AS
SELECT f.id, u.nickname, u.fullname, f.body, f.ts
FROM flipr.users u
JOIN flipr.flips f ON u.nickname = f.nickname;
COMMIT;
Use a "SELECT" statement in verify/userflips.sql again:
-- Verify flipr:userflips on exasol
SELECT id, nickname, fullname, body, ts
FROM flipr.userflips
WHERE FALSE;
ROLLBACK;
And of course, its "revert" script, revert/userflips.sql, should look something like:
-- Revert flipr:userflips from exasol
DROP VIEW flipr.userflips;
COMMIT;
Try em out!
> sqitch deploy
Deploying changes to flipr_test
+ flips ...... ok
+ userflips .. ok
Do we have the new table and view? Of course we do, they were verified. Still, have a
look:
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "select table_name from exa_all_tables where table_schema = 'FLIPR';"
TABLE_NAME
--------------------------------------------------------------------------------------------------------------------------------
USERS
FLIPS
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.userflips;"
COLUMN_NAME SQL_TYPE NULLABLE DISTRIBUTION_KEY
-------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- -------- ----------------
ID DECIMAL(18,0)
NICKNAME VARCHAR(64) UTF8
FULLNAME VARCHAR(256) UTF8
BODY VARCHAR(180) UTF8
TS TIMESTAMP WITH LOCAL TIME ZONE
And what's the status?
> sqitch status
# On database flipr_test
# Project: flipr
# Change: d1f998618fb863d93049a724fd0d2b49a29add86
# Name: userflips
# Deployed: 2014-09-04 17:51:21 -0700
# By: Marge N. OXVera <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
- userflips .. ok
- flips ...... ok
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.flips;"
Error: [42000] table or view FLIPR.FLIPS not found [line 1, column 10] (Session: 1582955242708359302)
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.userflips;"
Error: [42000] table or view FLIPR.USERFLIPS not found [line 1, column 10] (Session: 1582955248116468907)
Note the use of "@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 flips table and userflips view.'
[main c40f23f] Add flips table and userflips view.
7 files changed, 41 insertions(+), 0 deletions(-)
create mode 100644 deploy/flips.sql
create mode 100644 deploy/userflips.sql
create mode 100644 revert/flips.sql
create mode 100644 revert/userflips.sql
create mode 100644 verify/flips.sql
create mode 100644 verify/userflips.sql
> sqitch deploy
Deploying changes to flipr_test
+ flips ...... ok
+ userflips .. ok
> sqitch status
# On database flipr_test
# Project: flipr
# Change: d1f998618fb863d93049a724fd0d2b49a29add86
# Name: userflips
# Deployed: 2014-09-04 17:59:34 -0700
# By: Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
> sqitch verify
Verifying flipr_test
* appschema .. ok
* users ...... ok
* flips ...... ok
* userflips .. ok
Verify successful
Great, we're fully up-to-date!
Ship It!
Let's do a first release of our app. Let's call it "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 "userflips" with @v1.0.0-dev1
> git commit -am 'Tag the database with v1.0.0-dev1.'
[main b07ce3d] Tag the database with v1.0.0-dev1.
1 files changed, 1 insertions(+), 0 deletions(-)
> 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:
> sqitch deploy
Nothing to deploy (up-to-date)
> sqitch status
# On database flipr_test
# Project: flipr
# Change: d1f998618fb863d93049a724fd0d2b49a29add86
# Name: userflips
# Tag: @v1.0.0-dev1
# Deployed: 2014-09-04 17:59:34 -0700
# By: Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
Note the new "Tag" line in the output of "sqitch status": no new changes needed to be
deployed, but Sqitch did deploy the tag on the "userflips" change. Now let's bundle
everything up for release:
> sqitch bundle
Bundling into bundle
Writing config
Writing plan
Writing scripts
+ appschema
+ users
+ flips
+ userflips @v1.0.0-dev1
Now we can package the bundle directory and distribute it. When it gets installed
somewhere, users can use Sqitch to deploy to the database. Let's try deploying it to
another database. First, start up a new Docker instance on a different port:
> docker run --detach --privileged --stop-timeout 120 -p 127.0.0.1:9999:8888 exasol/docker-db:latest
Now you should be able to deploy to the new database:
> cd bundle
> sqitch deploy 'db:exasol://sys:exasol@localhost:9999/?Driver=Exasol'
Adding registry tables to db:exasol://sys:@localhost:9999/?Driver=Exasol
Deploying changes to db:exasol://sys:@localhost:9999/?Driver=Exasol
+ appschema ............... ok
+ users ................... ok
+ flips ................... ok
+ userflips @v1.0.0-dev1 .. ok
Notice how the tag on "userflips" now appears in the deploy output. Nice, eh? Now,
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
Making a Hash of Things
Now that we've got the basics of the app done, let's add a feature. Gotta track the
hashtags associated with flips, right? Let's add a table for them. But 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 hashtags
Switched to a new branch 'hashtags'
Now we can add a new change to create a table for hashtags.
> sqitch add hashtags --requires flips -n 'Adds table for storing hashtags.'
Created deploy/hashtags.sql
Created revert/hashtags.sql
Created verify/hashtags.sql
Added "hashtags [appschema flips]" to sqitch.plan
You know the drill by now. Add this to deploy/hashtags.sql
CREATE TABLE flipr.hashtags (
flip_id INTEGER NOT NULL REFERENCES flipr.flips(id),
hashtag VARCHAR(128) NOT NULL,
PRIMARY KEY (flip_id, hashtag)
);
Again, select from the table in verify/hashtags.sql:
SELECT flip_id, hashtag FROM flipr.hashtags WHERE FALSE;
And drop it in revert/hashtags.sql
DROP TABLE flipr.hashtags;
And give it a whirl:
> sqitch deploy
Deploying changes to flipr_test
+ hashtags .. ok
Look good?
> sqitch status --show-tags
# On database flipr_test
# Project: flipr
# Change: fda6daef73e0ac12252bf6af5f259ccb207d4197
# Name: hashtags
# Deployed: 2014-09-05 10:46:20 -0700
# By: Marge N. OXVera <marge@example.com>
#
# Tag:
# @v1.0.0-dev1 - 2014-09-05 09:09:38 -0700 - Marge N. OXVera <marge@example.com>
#
Nothing to deploy (up-to-date)
Note the use of "--show-tags" to show all the deployed tags. Make sure we can revert, too:
> sqitch rebase -y --onto @HEAD^
Reverting changes to userflips @v1.0.0-dev1 from flipr_test
- hashtags .. ok
> sqitch deploy
Deploying changes to flipr_test
+ hashtags .. ok
Great! Now make it so:
> git add .
> git commit -m 'Add hashtags table.'
[hashtags d893e9c] Add hashtags table.
4 files changed, 18 insertions(+), 0 deletions(-)
create mode 100644 deploy/hashtags.sql
create mode 100644 revert/hashtags.sql
create mode 100644 verify/hashtags.sql
Good, we've finished this feature. Time to merge back into "main".
Emergency
Let's do it:
> git checkout main
Switched to branch 'main'
> git pull
Updating b07ce3d..05d3e5d
Fast-forward
deploy/lists.sql | 10 ++++++++++
revert/lists.sql | 3 +++
sqitch.plan | 2 ++
verify/lists.sql | 5 +++++
4 files changed, 20 insertions(+), 0 deletions(-)
create mode 100644 deploy/lists.sql
create mode 100644 revert/lists.sql
create mode 100644 verify/lists.sql
Hrm, that's interesting. Looks like someone made some changes to "main". They added list
support. Well, let's see what happens when we merge our changes.
> git merge --no-ff hashtags
Auto-merging sqitch.plan
CONFLICT (content): Merge conflict in sqitch.plan
Automatic merge failed; fix conflicts and then commit the result.
Oh, a conflict in sqitch.plan. Not too surprising, since both the merged "lists" branch
and our "hashtags" branch added changes to the plan. Let's try a different approach.
The truth is, we got lazy. Those changes when we pulled main from the origin should have
raised a red flag. It's considered a bad practice not to look at what's changed in "main"
before merging in a branch. What one should do is either:
• Rebase the hashtags branch from main before merging. This "rewinds" the branch
changes, pulls from "main", and then replays the changes back on top of the pulled
changes.
• Create a patch and apply that to main. This is the sort of thing you might have to do
if you're sending changes to another user, especially if the VCS is not Git.
So let's restore things to how they were at main:
> git reset --hard HEAD
HEAD is now at 05d3e5d Merge branch 'lists'
That throws out our botched merge. Now let's go back to our branch and rebase it on
"main":
> git checkout hashtags
Switched to branch 'hashtags'
> git rebase main
First, rewinding head to replay your work on top of it...
Applying: Add hashtags table.
Using index info to reconstruct a base tree...
<stdin>:16: new blank line at EOF.
+
warning: 1 line adds whitespace errors.
Falling back to patching base and 3-way merge...
Auto-merging sqitch.plan
CONFLICT (content): Merge conflict in sqitch.plan
Failed to merge in the changes.
Patch failed at 0001 Add hashtags table.
When you have resolved this problem run "git rebase --continue".
If you would prefer to skip this patch, instead run "git rebase --skip".
To restore the original branch and stop rebasing run "git rebase --abort".
Oy, that's kind of a pain. It seems like no matter what we do, we'll need to resolve
conflicts in that file. Except in Git. Fortunately for us, we can tell Git to resolve
conflicts in sqitch.plan differently. Because we only ever append lines to the file, we
can have it use the "union" merge driver, which, according to its docs <https://git-
scm.com/docs/gitattributes#_built-in_merge_drivers>:
Run 3-way file level merge for text files, but take lines from both versions, instead
of leaving conflict markers. This tends to leave the added lines in the resulting file
in random order and the user should verify the result. Do not use this if you do not
understand the implications.
This has the effect of appending lines from all the merging files, which is exactly what
we need. So let's give it a try. First, back out the botched rebase:
> git rebase --abort
HEAD is now at d893e9c Add hashtags table.
Now add the union merge driver to .gitattributes for sqitch.plan and rebase again:
> echo sqitch.plan merge=union > .gitattributes
> git rebase main
First, rewinding head to replay your work on top of it...
Applying: Add hashtags table.
Using index info to reconstruct a base tree...
<stdin>:16: new blank line at EOF.
+
warning: 1 line adds whitespace errors.
Falling back to patching base and 3-way merge...
Auto-merging sqitch.plan
Ah, that looks a bit better. Let's have a look at the plan:
> cat sqitch.plan
%syntax-version=1.0.0
%project=flipr
%uri=https://github.com/sqitchers/sqitch-exasol-intro/
appschema 2014-09-04T18:40:34Z Marge N. OXVera <marge@example.com> # Add schema for all flipr objects.
users [appschema] 2014-09-04T23:40:15Z Marge N. OXVera <marge@example.com> # Creates table to track our users.
flips [appschema users] 2014-09-05T00:16:58Z Marge N. OXVera <marge@example.com> # Adds table for storing flips.
userflips [appschema users flips] 2014-09-05T00:18:43Z Marge N. OXVera <marge@example.com> # Creates the userflips view.
@v1.0.0-dev1 2014-09-05T16:04:48Z Marge N. OXVera <marge@example.com> # Tag v1.0.0-dev1.
lists [appschema users] 2014-09-05T17:33:43Z Marge N. OXVera <marge@example.com> # Adds table for storing lists.
hashtags [appschema flips] 2014-09-05T17:39:53Z Marge N. OXVera <marge@example.com> # Adds table for storing hashtags.
Note that it has appended the changes from the merged "lists" branch, and then merged the
changes from our "hashtags" branch. Test it to make sure it works as expected:
> sqitch rebase -y
Reverting all changes from flipr_test
- hashtags ................ ok
- userflips @v1.0.0-dev1 .. ok
- flips ................... ok
- users ................... ok
- appschema ............... ok
Deploying changes to flipr_test
+ appschema ............... ok
+ users ................... ok
+ flips ................... ok
+ userflips @v1.0.0-dev1 .. ok
+ lists ................... ok
+ hashtags ................ ok
Note the use of "rebase", which combines a "revert" and a "deploy" into a single command.
Handy, right? It correctly reverted our changes, and then deployed them all again in the
proper order. So let's commit .gitattributes; seems worthwhile to keep that change:
> git add .
> git commit -m 'Add `.gitattributes` with union merge for `sqitch.plan`.'
[hashtags 2f065a3] Add `.gitattributes` with union merge for `sqitch.plan`.
1 files changed, 1 insertions(+), 0 deletions(-)
create mode 100644 .gitattributes
Merges Mastered
And now, finally, we can merge into "main":
> git checkout main
Switched to branch 'main'
> git merge --no-ff hashtags -m "Merge branch 'hashtags'"
Merge made by recursive.
.gitattributes | 1 +
deploy/hashtags.sql | 10 ++++++++++
revert/hashtags.sql | 3 +++
sqitch.plan | 1 +
verify/hashtags.sql | 3 +++
5 files changed, 18 insertions(+), 0 deletions(-)
create mode 100644 .gitattributes
create mode 100644 deploy/hashtags.sql
create mode 100644 revert/hashtags.sql
create mode 100644 verify/hashtags.sql
And double-check our work:
> cat sqitch.plan
%syntax-version=1.0.0
%project=flipr
%uri=https://github.com/sqitchers/sqitch-exasol-intro/
appschema 2014-09-04T18:40:34Z Marge N. OXVera <marge@example.com> # Add schema for all flipr objects.
users [appschema] 2014-09-04T23:40:15Z Marge N. OXVera <marge@example.com> # Creates table to track our users.
flips [appschema users] 2014-09-05T00:16:58Z Marge N. OXVera <marge@example.com> # Adds table for storing flips.
userflips [appschema users flips] 2014-09-05T00:18:43Z Marge N. OXVera <marge@example.com> # Creates the userflips view.
@v1.0.0-dev1 2014-09-05T16:04:48Z Marge N. OXVera <marge@example.com> # Tag v1.0.0-dev1.
lists [appschema users] 2014-09-05T17:33:43Z Marge N. OXVera <marge@example.com> # Adds table for storing lists.
hashtags [appschema flips] 2014-09-05T17:39:53Z Marge N. OXVera <marge@example.com> # Adds table for storing hashtags.
Much much better, a nice clean main now. And because it is now identical to the "hashtags"
branch, we can just carry on. Go ahead and tag it, bundle, and release:
> sqitch tag v1.0.0-dev2 -n 'Tag v1.0.0-dev2.'
Tagged "hashtags" with @v1.0.0-dev2
> git commit -am 'Tag the database with v1.0.0-dev2.'
[main 8a6a73b] Tag the database with v1.0.0-dev2.
1 files changed, 1 insertions(+), 0 deletions(-)
> git tag v1.0.0-dev2 -am 'Tag v1.0.0-dev2'
> sqitch bundle --dest-dir flipr-1.0.0-dev2
Bundling into flipr-1.0.0-dev2
Writing config
Writing plan
Writing scripts
+ appschema
+ users
+ flips
+ userflips @v1.0.0-dev1
+ lists
+ hashtags @v1.0.0-dev2
Note the use of the "--dest-dir" option to "sqitch bundle". Just a nicer way to create the
top-level directory name so we don't have to rename it from bundle.
In Place Changes
Well, some folks have been testing the "1.0.0-dev2" release and have demanded that Twitter
user links be added to Flipr pages. Why anyone would want to include social network links
in an anti-social networking app is beyond us programmers, but we're just the plumbers,
right? Gotta go with what Product demands. The upshot is that we need to update the
"userflips" view, which is used for the feature in question, to include the Twitter user
names.
Normally, modifying views in database changes is a PITA
<https://www.urbandictionary.com/define.php?term=pita>. You have to make changes like
these:
1. Copy deploy/userflips.sql to deploy/userflips_twitter.sql.
2. Edit deploy/userflips_twitter.sql to drop and re-create the view with the "twitter"
column to the view.
3. Copy deploy/userflips.sql to revert/userflips_twitter.sql. Yes, copy the original
change script to the new revert change.
4. Add a "DROP VIEW" statement to revert/userflips_twitter.sql.
5. Copy verify/userflips.sql to verify/userflips_twitter.sql.
6. Modify verify/userflips_twitter.sql to include a check for the "twiter" column.
7. Test the changes to make sure you can deploy and revert the "userflips_twitter"
change.
But you can have Sqitch do most of the work for you. The only requirement is that a tag
appear between the two instances of a change we want to modify. In general, you're going
to make a change like this after a release, which you've tagged anyway, right? Well we
have, with "@v1.0.0-dev2" added in the previous section. With that, we can let Sqitch do
most of the hard work for us, thanks to the "rework" command, which is similar to "add":
> sqitch rework userflips -n 'Adds userflips.twitter.'
Added "userflips [userflips@v1.0.0-dev2]" to sqitch.plan.
Modify these files as appropriate:
* deploy/userflips.sql
* revert/userflips.sql
* verify/userflips.sql
Oh, so we can edit those files in place. Nice! How does Sqitch do it? Well, in point of
fact, it has copied the files to stand in for the previous instance of the "userflips"
change, which we can see via "git status":
> git status
# On branch main
# Changed but not updated:
# (use "git add <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# modified: revert/userflips.sql
# modified: sqitch.plan
#
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# deploy/userflips@v1.0.0-dev2.sql
# revert/userflips@v1.0.0-dev2.sql
# verify/userflips@v1.0.0-dev2.sql
no changes added to commit (use "git add" and/or "git commit -a")
The "untracked files" part of the output is the first thing to notice. They're all named
"userflips@v1.0.0-dev2.sql". What that means is: "the "userflips" change as it was
implemented as of the "@v1.0.0-dev2" tag." These are copies of the original scripts, and
thereafter Sqitch will find them when it needs to run scripts for the first instance of
the "userflips" change. As such, it's important not to change them again. But hey, if
you're reworking the change, you shouldn't need to.
The other thing to notice is that revert/userflips.sql has changed. Sqitch replaced it
with the original deploy script. As of now, deploy/userflips.sql and revert/userflips.sql
are identical. This is on the assumption that the deploy script will be changed (we're
reworking it, remember?), and that the revert script should actually change things back to
how they were before. Of course, the original deploy script may not be idempotent
<https://en.wikipedia.org/wiki/Idempotence> -- that is, able to be applied multiple times
without changing the result beyond the initial application. If it's not, you will likely
need to modify it so that it properly restores things to how they were after the original
deploy script was deployed. Or, more simply, it should revert changes back to how they
were as-of the deployment of deploy/userflips@v1.0.0-dev2.sql.
Fortunately, our view deploy scripts are already idempotent, thanks to the use of the "OR
REPLACE" expression. No matter how many times a deployment script is run, the end result
will be the same instance of the view, with no duplicates or errors.
As a result, there is no need to explicitly add changes. So go ahead. Modify the script to
add the "twitter" column to the view. Make this change to deploy/userflips.sql:
@@ -4,6 +4,6 @@
-- requires: flips
CREATE OR REPLACE VIEW flipr.userflips AS
-SELECT f.id, u.nickname, u.fullname, f.body, f.timestamp
+SELECT f.id, u.nickname, u.fullname, u.twitter, f.body, f.timestamp
FROM flipr.users u
JOIN flipr.flips f ON u.nickname = f.nickname;
Next, modify verify/userflips.sql to check for the "twitter" column. Here's the diff:
@@ -1,6 +1,6 @@
-- Verify flipr:userflips on exasol
-SELECT id, nickname, fullname, body, timestamp
+SELECT id, nickname, fullname, twitter, body, timestamp
FROM flipr.userflips
WHERE FALSE;
Now try a deployment:
> sqitch deploy
Deploying changes to flipr_test
+ userflips .. ok
So, are the changes deployed?
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.userflips;"
COLUMN_NAME SQL_TYPE NULLABLE DISTRIBUTION_KEY
-------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- -------- ----------------
ID DECIMAL(18,0)
NICKNAME VARCHAR(64) UTF8
FULLNAME VARCHAR(256) UTF8
TWITTER VARCHAR(256) UTF8
BODY VARCHAR(180) UTF8
TS TIMESTAMP WITH LOCAL TIME ZONE
Awesome, the view now includes the "twitter" column. But can we revert?
> sqitch revert --to @HEAD^ -y
Reverting changes to hashtags @v1.0.0-dev2 from flipr_test
- userflips .. ok
Did that work, is the "twitter" column gone?
> exaplus -q -u sys -p exasol -c localhost:8563 -sql "describe flipr.userflips;"
COLUMN_NAME SQL_TYPE NULLABLE DISTRIBUTION_KEY
-------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- -------- ----------------
ID DECIMAL(18,0)
NICKNAME VARCHAR(64) UTF8
FULLNAME VARCHAR(256) UTF8
BODY VARCHAR(180) UTF8
TS TIMESTAMP WITH LOCAL TIME ZONE
Yes, it works! Sqitch properly finds the original instances of these changes in the new
script files that include tags.
Excellent. Let's go ahead and commit these changes:
> git add .
> git commit -m 'Add the twitter column to the userflips view.'
[main 95d6dd0] Add the twitter column to the userflips view.
7 files changed, 30 insertions(+), 4 deletions(-)
create mode 100644 deploy/userflips@v1.0.0-dev2.sql
create mode 100644 revert/userflips@v1.0.0-dev2.sql
create mode 100644 verify/userflips@v1.0.0-dev2.sql
More to Come
Sqitch is a work in progress. Better integration with version control systems is planned
to make managing idempotent reworkings even easier. Stay tuned.
Author
David E. Wheeler <david@justatheory.com>
License
Copyright (c) 2012-2022 iovation Inc., David E. Wheeler
Permission is hereby granted, free of charge, to any person obtaining a copy of this
software and associated documentation files (the "Software"), to deal in the Software
without restriction, including without limitation the rights to use, copy, modify, merge,
publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons
to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or
substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE
FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.