# NAME App::Env - manage application specific environments # VERSION version 1.00 # SYNOPSIS # import environment from application1 then application2 into current # environment use App::Env ( $application1, $application2, \%opts ); # import an environment at your leisure use App::Env; App::Env::import( $application, \%opts ); # set defaults use App::Env ( \%defaults ) App::Env::config( %defaults ); # retrieve an environment but don't import it $env = App::Env->new( $application, \%opts ); # execute a command in that environment; just as a convenience $env->system( $command ); # exec a command in that environment; just as a convenience $env->exec( $command ); # oh bother, just import the environment $env->import; # cache this environment as the default for $application $env->cache( 1 ); # uncache this environment if it is the default for $application $env->cache( 0 ); # generate a string compatible with the *NIX env command $envstr = $env->str( \%opts ); # or, stringify it for (mostly) the same result system( 'env -i $env command' ); # pretend it's a hash; read only, though %ENV = %$env; # DESCRIPTION **App::Env** presents a uniform interface to initializing environments for applications which require special environments. **App::Env** only handles the loading, merging, and caching of environments; it does not create them. That is done within modules for each application suite (e.g. **App::Env::MyApp**). **App::Env** ships with two such modules, **App::Env::Null** which simply returns a snapshot of the current environment, and **App::Env::Example**, which provides example code for creating an application specific environment. **App::Env** is probably most useful in situations where a Perl program must invoke multiple applications each of which may require an environment different and possibly incompatible from the others. The simplified interface it provides makes it useful even in less complicated situations. ## Initializing Application Environments As mentioned above, **App::Env** does not itself provide the environments for applications; it relies upon application specific Perl modules to do so. Such modules must provide an **envs()** function which should return a hash reference containing the environment. Application specific options (e.g. version) may be passed to the module. See **App::Env::Example** for information on how to write such modules. ## Managing Environments In the simplest usage, **App::Env** can merge (`import`) the application's environment directly into the current environment. For situations where multiple incompatible environments are required, it can encapsulate those as objects with convenience methods to easily run applications within those environments. ## Environment Caching Environments are (by default) cached to improve performance; the default cache id is generated from the name of the Perl module which creates the environment and the options passed to it. signature. When a environment is requested its signature is compared against those stored in the cache and if matched, the associated cached environment is returned. The cache id is (by default) generated from the full module name (beginning with `App::Env` and including the optional site path -- see ["Site Specific Contexts"](#site-specific-contexts)) and the contents of the **AppOpts** hash passed to the module. If the **AppOpts** hash is empty, the id is just the module name. The cache id may be explicitly specified with the `CacheID` option. If `CacheID` is set to the string `AppID` the cache id is set to the full module name, ignoring the contents of **AppOpts**. This is useful if an application wishes to load an environment using special options but make it available under the more generic cache id. To prevent caching, use the `Cache` option. It doesn't prevent **App::Env** from _retrieving_ an existing cached environment -- to do that, use the `Force` option, which will result in a freshly generated environment. To retrieve a cached environment using its cache id use the **retrieve()** function. If multiple applications are loaded via a single call to **import** or **new** the applications will be loaded incrementally in the order specified. In order to ensure a properly merged environment the applications will be loaded freshly (any caches will be ignored) and the merged environment will be cached. The cache id will by default be generated from all of the names of the environment modules invoked; again, this can be overridden using the **CacheID** option. ## Application Aliases **App::Env** performs a case-insensitive search for application modules. For example, if the application module is named **App::Env::CIAO**, a request for `ciao` will resolve to it. Explicit aliases are also possible. A module should be created for each alias with the single class method **alias** which should return the name of the original application. For example, to make `App3` be an alias for `App1` create the following `App3.pm` module: package App::Env::App3; sub alias { return 'App1' }; 1; The aliased environment can provide presets for **AppOpts** by returning a hash as well as the application name: package App::Env::ciao34; sub alias { return 'CIAO', { Version => 3.4 } }; 1; These will be merged with any `AppOpts` passed in via **import()**, with the latter taking precedence. ## Site Specific Contexts In some situations an application's environment will depend upon which host or network it is executed on. In such instances **App::Env** provides a means for loading an alternate application module. It does this by loading the first existent module from the following set of module names: App::Env::$SITE::$app App::Env::$app The default value for `$SITE` is determined when `App::Env` is first loaded. If the environment variable `APP_ENV_SITE` exists it is set to that, otherwise if the `App::Env::Site` module exists, that is loaded. It should set the `APP_ENV_SITE` variable. After this, modifications to `APP_ENV_SITE` are ignored. The default value may be overridden via the [Site](https://metacpan.org/pod/Site) option passed to the class **import()** function or the **new()** object constructor. Take as an example the situation where an application's environment is stored in `/usr/local/myapp/setup` on one host and `/opt/local/myapp/setup` on another. One could include logic in a single `App::Env::myapp` module which would recognize which file is appropriate. If there are multiple applications, this gets messy. A cleaner method is to have separate site-specific modules (e.g. `App::Env::LAN1::myapp` and `App::Env::LAN2::myapp`), and switch between them based upon the **APP\_ENV\_SITE** environment variable. The logic for setting that variable might be encoded in an **App::Env::Site** module to transparently automate things: package App::Env::Site; my %LAN1 = map { ( $_ => 1 ) } qw( sneezy breezy queasy ); my %LAN2 = map { ( $_ => 1 ) } qw( dopey mopey ropey ); use Sys::Hostname; if ( $LAN1{hostname()} ) { $ENV{APP_ENV_SITE} = 'LAN1'; } elsif ( $LAN2{hostname()} ) { $ENV{APP_ENV_SITE} = 'LAN2'; } 1; ## The Null Environment **App::Env** provides the `null` environment, which simply returns a snapshot of the current environment. This may be useful to provide fall-backs in case an application specific environment was not found, but the code should fallback to using the existing environment. $env = eval { App::Env->new( "MyApp" ) } \ // App::Env->new( "null", { Force => 1, Cache => 0 } ); As the `null` environment is a _snapshot_ of the current environment, if future `null` environments should reflect the environment at the time they are constructed, C"null" environments should not be cached (e.g. `Cache => 0`). The `Force => 1` option is specified to ensure that the environment is not being read from cache, just in case a prior `null` environment was inadvertently cached. # INTERFACE **App::Env** may be used to directly import an application's environment into the current environment, in which case the non-object oriented interface will suffice. For more complicated uses, the object oriented interface allows for manipulating multiple separate environments. ## Using **App::Env** without objects Application environments may be imported into the current environment either when loading **App::Env** or via the **App::Env::import()** function. - import use App::Env ( $application, \%options ); use App::Env ( @applications, \%shared_options ); App::Env::import( $application, \%options ); App::Env::import( @applications, \%shared_options ); Import the specified applications. Options may be applied to specific applications by grouping application names and option hashes in arrays: use App::Env ( [ 'app1', \%app1_options ], [ 'app2', \%app2_options ], \%shared_options ); App::Env::import( [ 'app1', \%app1_options ], [ 'app2', \%app2_options ], \%shared_options ); Shared (or default) values for options may be specified in a hash passed as the last argument. The available options are listed below. Not all options may be shared; these are noted. - AppOpts _hashref_ This is a hash of options to pass to the `App::Env::` module. Their meanings are application specific. This option may not be shared. - Force _boolean_ Don't use the cached environment for this application. - Site Specify a site. See ["Application Environments"](#application-environments) for more information - Cache _boolean_ Cache (or don't cache) the environment. By default it is cached. If multiple environments are loaded the _combination_ is also cached. - CacheID A unique name for the environment. See ["Environment Caching"](#environment-caching) for more information. When used as a shared option for multiple applications, this will be used to identify the merged environment. If set to the string `AppID`, the full module name will be used as the cache id (ignoring the contents of the **AppOpts** option hash). - SysFatal _boolean_ If true, the **system**, **qexec**, and **capture** object methods will throw an exception if the passed command exits with a non-zero error. - Temp _boolean_ If true, and the requested environment does not exist in the cache, create it but do not cache it (this overrides the **Cache** option). If the requested environment does exist in the cache, return an non-cached clone of it. The following options are updated in the cloned environment: SysFatal - retrieve $env = App::Env::retrieve( $cacheid ); Retrieve the environment with the given cache id, or undefined if it doesn't exist. ## Managing Environments - config App::Env::config( %Defaults ); Configure default options for environments. See ["Changing Default Option Values"](#changing-default-option-values) for more information. - uncache App::Env::uncache( App => $app, [ Site => $site ] ) App::Env::uncache( CacheID => $cacheid ) Delete the cache entry for the given application. If `Site` is not specified, the site is determined as specified in ["Site Specific Contexts"](#site-specific-contexts). It is currently _not_ possible to use this interface to explicitly uncache multi-application environments if they have not been given a unique cache id. It is possible using **App::Env** objects. The available options are: - App The application name. This may not be specified if **CacheID** is specified. - Site If the **Site** option was used when first loading the environment, it must be specified here in order to delete the correct cache entry. Do not specify this option if **CacheID** is specified. - CacheID If the **CacheID** option was used to provide a cache key for the cache entry, this must be specified here. Do not specify this option if **App** or **Site** are specified. - All If true uncache all of the cached environments. ## Using **App::Env** objects **App::Env** objects give greater flexibility when dealing with multiple applications with incompatible environments. ### Constructors - new $env = App::Env->new( ... ) **new** takes the same arguments as **App::Env::import** and returns an **App::Env** object. It does not modify the environment. - clone $clone = $app->clone( \%opts ); Clone an existing environment. The available options are `CacheID`, `Cache`, `SysFatal` (see the documentation for the **import** function). The cloned environment is by default not cached. If caching is requested and a cache id is not provided, a unique id is created -- it will _not_ be the same as that of the original environment. This generated cache id is not based on a signature of the environment, so this environment will effectively not be automatically reused when a similar environment is requested via the **new** constructor (see ["Environment Caching"](#environment-caching)). ### Overloaded operators **App::Env** overloads the %{} and "" operators. When dereferenced as a hash an **App::Env** object returns a hash of the environmental variables: %ENV = %$env; When interpolated in a string, it is replaced with a string suitable for use with the \*NIX **env** command; see the **str()** method below for its format. ### Methods - cache $env->cache( $cache_state ); If `$cache_state` is true, cache this environment using the object's cache id. If `$cache_state` is false and this environment is being cached, delete the cache. Note that only the original **App::Env** object which cached the environment may delete it. Objects which reuse existing, cached, environments cannot. - cacheid $cacheid = $env->cacheid; Returns the cache id for this environment. - env # return a hashref of the entire environment (similar to %{$env}) $hashref = $env->env( ); # return the value of a given variable in the environment $value = $env->env( $variable_name ) # return an array of values of particular variables. # names should be strings @values = $env->env( @variable_names ); # match variable names and return a hashref $hashref = $env->env( @match_specifications ); # exclude specific variables $hashref = $env->env( { Exclude => $match_spec } ); $hashref = $env->env( { Exclude => \@match_specs } ); $hashref = $env->env( @match_specs, { Exclude => $match_spec } ); $hashref = $env->env( @match_specs, { Exclude => \@match_specs } ); Return all or parts of the environment. What is returned depends upon the type of argument and which of the following contexts matches: 1. If called with no arguments (or just an **Exclude** option, as discussed below) return a hashref containing the environment. 2. If called in a scalar context and passed a single variable name (which must be a string) return the value for that variable, or _undef_ if it is not in the environment. 3. If called in a list context and passed a list of variable names (which must be strings) return an array of values for those variables (_undef_ for those not in the environment). 4. If called in a scalar context and passed one or more _match specifications_, return a hashref containing the subset of the environment which matches. The `Exclude` option (see below) may be present. A _match specification_ may be a string, (for an exact match of a variable name), a regular expression created with the **qr** operator, or a subroutine reference. The subroutine will be passed two arguments, the variable name and its value, and should return true if the variable should be excluded, false otherwise. To avoid mistaking this context for context 1 if the _match specification_ is a single string, enclose it in an array, e.g. # this is context 1 $value = $env->env( $variable_name ); # this is context 3 $hash = $env->env( [ $variable_name ] ); Variable names may be excluded from the list by passing a hash with the key `Exclude` as the last argument (valid only in contexts 0 and 3). The value is either a scalar or an arrayref composed of match specifications (as an arrayref) as described in context 3. - setenv # set an environmental variable $env->setenv( $var, $value ); # delete an environmetal variable $env->setenv( $var ); If `$value` is present, assign it to the named environmental variable. If it is not present, delete the variable. **Note:** If the environment refers to a cached environment, this will affect all instances of the environment which share the cache. - module $module = $env->module; This returns the name of the module which was used to load the environment. If multiple modules were used, the names are concatenated, separated by the `$;` (subscript separator) character. - str $envstr = $env->str( @match_specifications, \%options ); This function returns a string which may be used with the \*NIX **env** command to set the environment. The string contains space separated `var=value` pairs, with shell magic characters escaped. The environment may be pared down by passing _match specifications_ and an `Exclude` option; see the documentation for the **env** method, context 3, for more information. Because the **TERMCAP** environment variable is often riddled with escape characters, which are not always handled well by shells, the **TERMCAP** variable is _always_ excluded unless it is explicitly included via an exact variable name match specification. For example, $envstr = $env->str( qr/.*/, 'TERMCAP ); is the only means of getting all of the environment returned. - system $env->system( $command, @args ); This runs the passed command in the environment defined by **$env**. It has the same argument and returned value convention as the core Perl **system** command. If the **SysFatal** flag is set for this environment, **IPC::System::Simple::system** is called, which will cause this method to throw an exception if the command returned a non-zero exit value. It also avoid invoking a shell to run the command if possible. - exec $env->exec( $command, @args ); This execs the passed command in the environment defined by **$env**. It has the same argument and returned value convention as the core Perl **exec** command. - qexec $output = $env->qexec( $command, @args ); @lines = $env->qexec( $command, @args ); This acts like the **qx{}** Perl operator. It executes the passed command in the environment defined by **$env** and returns its (standard) output. If called in a list context the output is split into lines. If the **SysFatal** flag is set for this environment, **IPC::System::Simple::capture** is called, which will cause this method to throw an exception if the command returned a non-zero exit value. It also avoid invoking a shell to run the command if possible. - capture $stdout = $env->capture( $command, @args ); ($stdout, $stderr) = $env->capture( $command, @args ); Execute the passed command in the environment defined by **$env** and returns content of its standard output and (optionally) standard error streams. If the **SysFatal** flag is set for this environment, **IPC::System::Simple::capture** is called, which will cause this method to throw an exception if the command returned a non-zero exit value. It also avoid invoking a shell to run the command if possible. - which $path = $env->which( $command ); @paths = $env->which( $command ); Return the path (or paths in list mode) of the passed command using [File::Which](https://metacpan.org/pod/File%3A%3AWhich). It returns `undef` or an empty list if the command is not found. ## Changing Default Option Values Default values for some options may be changed via any of the following: - Passing a hashref as the only argument when initially importing the package: use App::Env \%Default; - Calling the **config** function: App::Env::config( %Default ); The following options may have their default values changed: Force Cache Site SysFatal # EXAMPLE USAGE ## A single application This is the simplest case. If you don't care if you "pollute" the current environment, then simply use App::Env qw( ApplicationName ); ## A single application with options If the **CIAO** environment module provides a `Version` option: use App::Env ( 'CIAO', { AppOpts => { Version => 3.4 } } ); ## Two compatible applications If two applications can share an environment, and you don't mind changing the current environment; use App::Env qw( Application1 Application2 ); If you need to preserve the environment you need to be a little more circumspect. $env = App::Env->new( qw( Application1 Application 2 ) ); $env->system( $command1, @args ); $env->system( $command2, @args ); or even $env->system( "$command1 | $command2" ); Or, { local %ENV = %$env; system( $command1); } if you prefer not to use the **system** method. ## Two incompatible applications If two applications can't share the environment, you'll need to load them separately: $env1 = App::Env->new( 'Application1' ); $env2 = App::Env->new( 'Application2' ); $env1->system( $command1 ); $env2->system( $command2 ); Things are trickier if you need to construct a pipeline. That's where the \*NIX **env** command and **App::Env** object stringification come into play: system( "env -i $env1 $command1 | env -i $env2 $command2" ); This hopefully won't overfill the shell's command buffer. If you need to specify only parts of the environment, use the **str** method to explicitly create the arguments to the **env** command. ## Localizing changes to an environment In some contexts an environment must be customized but the changes shouldn't propagate into the (possibly) cached version. A good example of this is in sandboxing functions which may manipulate an environment. The **new()** constructor doesn't indicate whether an environment was freshly constructed or pulled from cache, so the user can't tell if manipulating it will affect other code paths. One way around this is to force construction of a fresh environment using the `Force` option and turning off caching via the `Cache` option. This guarantees isolation but is inefficient (if a compatible environment is cached it won't be used) and any tweaks made to the environment by the application are not seen. Instead, use the `Temp` option; this will either create a new environment if none exists or clone an existing one. In either case the result won't be cached and any changes will be localized. # SUPPORT ## Bugs Please report any bugs or feature requests to bug-app-env@rt.cpan.org or through the web interface at: https://rt.cpan.org/Public/Dist/Display.html?Name=App-Env ## Source Source is available at https://gitlab.com/djerius/App-Env and may be cloned from https://gitlab.com/djerius/App-Env.git # SEE ALSO Please see those modules/websites for more information related to this module. - [**appexec**](https://metacpan.org/pod/appexec) # AUTHOR Diab Jerius # COPYRIGHT AND LICENSE This software is Copyright (c) 2018 by Smithsonian Astrophysical Observatory. This is free software, licensed under: The GNU General Public License, Version 3, June 2007