App::migrate - upgrade / downgrade project
This document describes App::migrate version v0.2.6
use App::migrate; my $migrate = App::migrate->new() $migrate = $migrate->load($file) @paths = $migrate->find_paths($v_from => $v_to) say "versions: @{$_}" for @paths; @steps = $migrate->get_steps($paths[0]) for (@steps) { say "$_->{prev_version} ... $_->{next_version}"; if ($_->{type} eq 'VERSION' or $_->{type} eq 'RESTORE') { say "$_->{type} $_->{version}"; } else { say "$_->{type} $_->{cmd} @{$_->{args}}"; } } $migrate = $migrate->on(BACKUP => sub{ my $step=shift; return or die }); $migrate = $migrate->on(RESTORE => sub{ my $step=shift; return or die }); $migrate = $migrate->on(VERSION => sub{ my $step=shift; return or die }); $migrate = $migrate->on(error => sub{ my $step=shift; return or die }); $migrate->run($paths[0]);
If you're looking for command-line tool - see migrate. This module is actual implementation of that tool's functionality and you'll need it only if you're developing similar tool (like narada-install) to implement specifics of your project in single perl script instead of using several external scripts.
This module implements file format (see "SYNTAX") to describe sequence of upgrade and downgrade operations needed to migrate something between different versions, and API to analyse and run these operations.
The something mentioned above is usually some project, but it can be literally anything - OS configuration in /etc, or overall OS setup including installed packages, etc. - anything what has versions and need complex operations to upgrade/downgrade between these versions. For example, to migrate source code you can use VCS like Git or Mercurial, but they didn't support empty directories, file permissions (except executable), non-plain file types (FIFO, UNIX socket, etc.), xattr, ACL, configuration files which must differ on each site, and databases. So, if you need to migrate anything isn't supported by VCS - you can try this module/tool.
Sometimes it isn't possible to really downgrade because some data was lost while upgrade - to handle these situations you should provide a ways to create complete backup of your project and restore any project's version from these backups while downgrade (of course, restoring backups will result in losing new changes, so whenever possible it's better to do some extra work to provide a way to downgrade without losing any data).
Here is example how to run migration from version '1.1.8' to '1.2.3' of some project which uses even minor versions '1.0.x' and '1.2.x' for stable releases and odd minor versions '1.1.x' for unstable releases. The nearest common version between '1.1.8' and '1.2.3' is '1.0.42', which was the parent for both '1.1.x' and '1.2.x' branches, so we need to downgrade project from '1.1.8' to '1.0.42' first, and then upgrade from '1.0.42' to '1.2.3'. You'll need two *.migrate files, one which describe migrations from '1.0.42' (or earlier version) to '1.1.8', and another with migrations from '1.0.42' (or earlier) to '1.2.3'. For brevity let's not make any backups while migration.
*.migrate
my $migrate = App::migrate ->new ->load('1.1.8.migrate') ->load('1.2.3.migrate'); $migrate ->on(BACKUP => sub {}) ->run( $migrate->find_paths('1.1.8' => '1.2.3') );
$migrate = App::migrate->new;
Create and return new App::migrate object.
$migrate->load('path/to/migrate');
Load migration commands into $migrate object.
$migrate
You should load at least one file with migration commands before you can use "find_paths", "get_steps" or "run".
When loading multiple files, if they contain two adjoining 'VERSION' operations with same version values then migration commands between these two version values will be used from first loaded file containing these version values.
Will throw if given file's contents don't conform to "Specification" - this may be used to check file's syntax.
@paths = $migrate->find_paths($from_version => $to_version);
Find and return all possible paths to migrate between given versions.
If no paths found - return empty list. This may happens because you didn't loaded migrate files which contain required migrations or because there is no way to migrate between these versions (for example, if one of given versions is incorrect).
Multiple paths can be found, for example, when your project had some branches which was later merged.
Each found path returned as single ARRAYREF element in returned list. This ARRAYREF contains list of all intermediate versions, one by one, starting from $from_version and ending with $to_version.
$from_version
$to_version
For example, if our project have this version history:
1.0.0 | 1.0.42 / \ 1.1.0 1.2.0 | | 1.1.8 1.2.3 | \ | | \----| 1.1.9 1.2.4 | | 1.1.10 1.2.5
then you'll probably have these migrate files:
1.1.10.migrate 1.0.0->…->1.0.42->1.1.0->…->1.1.10 1.2.5.migrate 1.0.0->…->1.0.42->1.2.0->…->1.2.3->1.2.4->1.2.5 1.1.8-1.2.4.migrate 1.0.0->…->1.0.42->1.1.0->…->1.1.8->1.2.4
If you "load" files 1.2.5.migrate and 1.1.8-1.2.4.migrate and then call find_paths('1.0.42' => '1.2.5'), then it will return this list with two paths (in any order):
1.2.5.migrate
1.1.8-1.2.4.migrate
find_paths('1.0.42' => '1.2.5')
( ['1.0.42', '1.1.0', …, '1.1.8', '1.2.4', '1.2.5'], ['1.0.42', '1.2.0', …, '1.2.3', '1.2.4', '1.2.5'], )
@steps = $migrate->get_steps( \@versions );
Return list of all migration operations needed to migrate on path given in @versions.
@versions
For example, to get steps for first path returned by "find_paths":
@steps = $migrate->get_steps( $migrate->find_paths($from=>$to) );
Steps returned in order they'll be executed while "run" for this path. Each element in @steps is a HASHREF with these keys:
@steps
type => one of these values: 'VERSION', 'before_upgrade', 'upgrade', 'downgrade', 'after_downgrade', 'RESTORE' # these keys exists only if value of type key is one of: # VERSION, RESTORE version => version number # these keys exists only if value of type key is one of: # before_upgrade, upgrade, downgrade, after_downgrade cmd => command to run args => ARRAYREF of params for that command
Will throw if unable to return requested steps.
$migrate = $migrate->on(BACKUP => \&your_handler); $migrate = $migrate->on(RESTORE => \&your_handler); $migrate = $migrate->on(VERSION => \&your_handler); $migrate = $migrate->on(error => \&your_handler);
Set handler for given event.
All handlers will be called only by "run"; they will get single parameter - step HASHREF (BACKUP handler will get step in same format as RESTORE), see "get_steps" for details of that HASHREF contents. Also these handlers may use $ENV{MIGRATE_PREV_VERSION} and $ENV{MIGRATE_NEXT_VERSION} - see "run" for more details.
$ENV{MIGRATE_PREV_VERSION}
$ENV{MIGRATE_NEXT_VERSION}
Handler will be executed when project backup should be created: before starting any new migration, except next one after RESTORE.
If handler throws then 'error' handler will be executed.
Default handler will throw (because it doesn't know how to backup your project).
NOTE: If you'll use handler which doesn't really create and keep backups for all versions then it will be impossible to do RESTORE operation.
Handler will be executed when project should be restored from backup: when downgrading between versions which contain RESTORE operation or when migration fails.
Default handler will throw (because it doesn't know how to restore your project).
Handler will be executed after each successful migration.
Default handler does nothing.
Handler will be executed when one of commands executed while migration fails or when BACKUP, RESTORE or VERSION handlers throw.
If handler throws then try to restore version-before-migration (without calling error handler again if it throws too).
Default handler will run $SHELL (to let you manually fix errors) and throw if you $SHELL exit status != 0 (to let you choose what to do next - continue migration if you fixed error or interrupt migration to restore version-before-migration from backup).
$migrate->run( \@versions );
Will use "get_steps" to get steps for path given in @versions and execute them in order. Will also call handlers as described in "on".
Before executing each step will set $ENV{MIGRATE_PREV_VERSION} to current version (which it will migrate from) and $ENV{MIGRATE_NEXT_VERSION} to version it is trying to migrate to.
Syntax of this file was designed to accomplish several goals:
Be able to automatically make sure each 'upgrade' operation has corresponding 'downgrade' operation (so it won't be forget - but, of course, it's impossible to automatically check is 'downgrade' operation will correctly undo effect of 'upgrade' operation).
Thus custom file format is needed.
Make it easier to manually analyse is 'downgrade' operation looks correct for corresponding 'upgrade' operation.
Thus related 'upgrade' and 'downgrade' operations must go one right after another.
Make it obvious some version can't be downgraded and have to be restored from backup.
Thus RESTORE operation is named in upper case.
Given all these requirements try to make it simple and obvious to define migrate operations, without needs to write downgrade code for typical cases.
Thus it's possible to define macro to turn combination of upgrade/downgrade operations into one user-defined operation (no worries here: these macro doesn't support recursion, it isn't possible to redefine them, and they have lexical scope - from definition to the end of this file - so they won't really add complexity).
VERSION 0.0.0 # To upgrade from 0.0.0 to 0.1.0 we need to create new empty file and # empty directory. upgrade touch empty_file downgrade rm empty_file upgrade mkdir empty_dir downgrade rmdir empty_dir VERSION 0.1.0 # To upgrade from 0.1.0 to 0.2.0 we need to drop old database. This # change can't be undone, so only way to downgrade from 0.2.0 is to # restore 0.1.0 from backup. upgrade rm useless.db RESTORE VERSION 0.2.0 # To upgrade from 0.2.0 to 1.0.0 we need to run several commands, # and after downgrading we need to kill some background service. before_upgrade patch <0.2.0.patch >/dev/null chmod +x some_daemon downgrade patch -R <0.2.0.patch >/dev/null upgrade ./some_daemon & after_downgrade killall -9 some_daemon VERSION 1.0.0 # Let's define some lazy helpers: DEFINE2 only_upgrade upgrade downgrade true DEFINE2 mkdir upgrade mkdir "$@" downgrade rm -rf "$@" # ... and use it: only_upgrade echo "Just upgraded to $MIGRATE_NEXT_VERSION" VERSION 1.0.1 # another lazy macro (must be defined above in same file) mkdir dir1 dir2 VERSION 1.1.0
Recommended name for file with upgrade/downgrade operations is either migrate or <version>.migrate.
migrate
<version>.migrate
Each line in migrate file must be one of these:
line start with symbol "#"
For comments. Line is ignored.
line start with any non-space symbol, except "#"
Contain one or more elements separated by one or more space symbols: operation name (case-sensitive), zero or more params (any param may be quoted, params which contain one of 5 symbols "\\\"\t\r\n" must be quoted).
Quoted params must be surrounded by double-quote symbol, and any of mentioned above 5 symbols must be escaped like shown above.
line start with two spaces
Zero or more such lines after line with operation name form one more, multiline, extra param for that operation (first two spaces will be removed from start of each line before providing this param to operation).
Not all operations may have such multiline param.
empty line
If this line is between operations then it's ignored.
If this line is inside operation's multiline param - then that multiline param will include this empty line.
If you will need to include empty line at end of multiline param then you'll have to use line with two spaces instead.
While executing any commands two environment variables will be set: $MIGRATE_PREV_VERSION and $MIGRATE_NEXT_VERSION (first is always version we're migrating from, and second is always version we're migrating to - i.e. while downgrading $MIGRATE_NEXT_VERSION will be lower/older version than $MIGRATE_PREV_VERSION)
$MIGRATE_PREV_VERSION
$MIGRATE_NEXT_VERSION
All executed commands must complete without error, otherwise emergency shell will be started and user should either fix the error and exit from shell to continue migration, or exit 1 from shell to interrupt migration and restore previous-before-this-migration version from backup.
exit
exit 1
Must have exactly one param (version number). Some symbols are not allowed in version numbers: special (0x00-0x1F,0x7F), both slash, all three quotes, ?, * and space.
Multiline param not supported.
This is delimiter between sequences of migrate operations.
Each file must contain 'VERSION' operation before any migrate operations (i.e. before first 'VERSION' operation only 'DEFINE', 'DEFINE2' and 'DEFINE4' operations are allowed).
All operations after last 'VERSION' operation will be ignored.
These operations must be always used in pairs: first must be one of 'before_upgrade' or 'upgrade' operation, second must be one of 'downgrade' or 'after_downgrade' or 'RESTORE' operations.
These four operations may have zero or more params and optional multiline param. If they won't have any params at all they'll be processed like they have one (empty) multiline param.
Their params will be executed as a single shell command at different stages of migration process and in different order:
On each migration only commands between two nearest VERSION operations will be processed.
On upgrading (migrate forward from previous VERSION to next VERSION) will be executed all 'before_upgrade' operations in forward order then all 'upgrade' operations in forward order.
On downgrading (migrate backward from next VERSION to previous) will be executed all 'downgrade' operations in backward order, then all 'after_downgrade' operations in backward order.
Shell command to use will be:
If operation has one or more params - first param will become executed command name, other params will become command params.
If operation also has multiline param then it content will be saved into temporary file and name of that file will be added at end of command's params.
Else multiline param will be saved into temporary file (after shebang #!/path/to/bash -ex if first line of multiline param doesn't start with #!), which will be made executable and run without any params.
#!/path/to/bash -ex
#!
Doesn't support any params, neither usual nor multiline.
Can be used only after 'before_upgrade' or 'upgrade' operations.
When one or more 'RESTORE' operations are used between some 'VERSION' operations then all 'downgrade' and 'after_downgrade' operations between same 'VERSION' operations will be ignored and on downgrading previous version will be restored from backup.
This operation must have only one non-multiline param - name of defined macro. This name must not be same as one of existing operation names, both documented here or created by one of previous 'DEFINE' or 'DEFINE2' or 'DEFINE4' operations.
Next operation must be one of 'before_upgrade', 'upgrade', 'downgrade' or 'after_downgrade' - it will be substituted in place of all next operations matching name of this macro.
When substituting macro it may happens what both this macro definition have some normal params and multiline param, and substituted operation also have some it's own normal params and multiline param. All these params will be combined into single command and it params in this way:
If macro definition doesn't have any params - params of substituted operation will be handled as usually for 'upgrade' etc. operations.
If macro definition have some params - they will be handled as usually for 'upgrade' etc. operations, so we'll always get some command and optional params for it.
Next, all normal params of substituted command (if any) will be appended to that command params.
Next, if substituted command have multiline param then it will be saved to temporary file and name of that file will be appended to that command params.
Work similar to DEFINE, but require two next operations after it: first must be one of 'before_upgrade' or 'upgrade', and second must be one of 'downgrade' or 'after_downgrade'.
Params of both operations will be combined with params of substituted operation as explained above.
Work similar to DEFINE, but require four next operations after it: first must be 'before_upgrade', second - 'upgrade', third - 'downgrade', fourth - 'after_downgrade'.
Params of all four operations will be combined with params of substituted operation as explained above.
Please report any bugs or feature requests through the issue tracker at https://github.com/powerman/migrate/issues. You will be notified automatically of any progress on your issue.
This is open source software. The code repository is available for public review and contribution under the terms of the license. Feel free to fork the repository and submit pull requests.
https://github.com/powerman/migrate
git clone https://github.com/powerman/migrate.git
MetaCPAN Search
https://metacpan.org/search?q=App-migrate
CPAN Ratings
http://cpanratings.perl.org/dist/App-migrate
AnnoCPAN: Annotated CPAN documentation
http://annocpan.org/dist/App-migrate
CPAN Testers Matrix
http://matrix.cpantesters.org/?dist=App-migrate
CPANTS: A CPAN Testing Service (Kwalitee)
http://cpants.cpanauthors.org/dist/App-migrate
Alex Efros <powerman@cpan.org>
This software is Copyright (c) 2015- by Alex Efros <powerman@cpan.org>.
This is free software, licensed under:
The MIT (X11) License
To install App::migrate, copy and paste the appropriate command in to your terminal.
cpanm
cpanm App::migrate
CPAN shell
perl -MCPAN -e shell install App::migrate
For more information on module installation, please visit the detailed CPAN module installation guide.