Devel::PerlySense - IntelliSense for Perl
PerlySense is an IntelliSense style utility for editors.
Conveniently navigate and browse the code and documentation of your project and Perl installation.
Run tests and scripts and syntax check source with easy navigation to errors/warnings/failing tests.
Highlight syntax errors, warnings and Perl::Critic complaints in the source while editing.
C-o C-c -- Class Overview -- Show information about the Class at point or the current Class.
C-o C-d -- Smart Docs -- Show docs (POD/signature/etc) for the symbol (module/method/sub) at point. A doc hint is displayed in the message area (for methods and subs), or a new POD buffer is created (for modules).
C-o C-g -- Smart Go to -- Open file at proper location for module, method/sub declaration for the symbol (module/method/sub) at point. If no sub declaration is available (like for generated getters/setters), any appropriate POD is used instead.
C-o C-r -- Run file -- Run the current file using the Compilation mode and the settings appropriate for the source type (Test, Module, etc.). Highlight errors and jump to source with C-c C-c.
C-o m f -- Perl Module open File -- Open the source file of the module at point.
Flymake may be used to highlight syntax errors and warnings in the source while editing (continously or at every save).
There is no integraton with Vim available. Well, yet.
Someone may want to write it.
Any editor that is programmable and that can call a shell script could take advantage of at least some parts of PerlySense to implement something similar to the Emacs functionality. And most editors are programmable by the authors, if not by the users.
perly_sense create_project [--dir=DIR]
Create a PerlySense project in DIR (default is current dir).
If there is already a project.yml file, back it up with a datestamp first.
perly_sense process_project
Cache all modules in the project. (not implemented)
perly_sense process_inc
Cache all the modules in @INC.
This is a useful thing to do after installation (and after each upgrade), but it will take a wile so put it in the background and let it churn away at those modules.
perly_sense info
Display useful information about what the current project directory, user home directory, etc. is.
See the source of the "perly_sense" in bin script, or the t directory.
Install required modules from CPAN.
Make sure the Devel::PerlySense CPAN module is installed, it contains the required elisp files which will be loaded automatically with the following in your .emacs config file:
;; *** PerlySense Config *** ;; ** PerlySense ** ;; The PerlySense prefix key (unset only if needed) (global-unset-key "\C-o") (setq perly-sense-key-prefix "\C-o") ;; ** Flymake ** ;; Load flymake if t ;; Flymake must be installed. ;; It is included in Emacs 22, or available from ;; http://flymake.sourceforge.net/ ;; Put flymake.el somewhere in your load-path. (setq perly-sense-load-flymake t) ;; Note: more flymake config below, after loading PerlySense ;; *** PerlySense load (don't touch) *** (setq perly-sense-external-dir (shell-command-to-string "perly_sense external_dir")) (if (string-match "Devel.PerlySense.external" perly-sense-external-dir) (progn (message "PerlySense elisp files at (%s) according to perly_sense, loading..." perly-sense-external-dir) (setq load-path (cons (expand-file-name (format "%s/%s" perly-sense-external-dir "emacs") ) load-path)) (load "perly-sense") (if perly-sense-load-flymake (load "perly-sense-flymake")) ) (message "Could not identify PerlySense install dir. Is Devel::PerlySense installed properly? Does 'perly_sense external_dir' give you a proper directory? (%s)" perly-sense-external-dir) ) ;; ** Flymake Config ** ;; If you only want syntax check whenever you save, not continously (setq flymake-no-changes-timeout 9999) (setq flymake-start-syntax-check-on-newline nil) ;; Emacs named colors: http://www.geocities.com/kensanata/colors.html ;; These colors work fine with a white X11 background. They may not look ;; that great on a console with the default color scheme. (set-face-background 'flymake-errline "antique white") (set-face-background 'flymake-warnline "lavender") ;; *** PerlySense End ***
The load path is handled automatically by asking "perly_sense external_dir" where the elisp source was installed (that way, Devel::PerlySense and the elisp is always in sync).
The most important config you can change is the prefix key.
The default, \C-o, seemed to have a rater low useful-to-keystroke ratio and so was a strong candidate for stealing for this much more important purpose :)
If you want to use flymake to do background syntax and Perl::Critic checks, set perly-sense-load-flymake to t (this is a very nifty thing, so yes you want to do this) and configure the colors to your liking.
Note: This also needs to be enabled on a per-project basis.
C-o C-d is the "Smart docs" command. It brings up documentation for what's at point.
Put the cursor on the "method" word of a $self->method call and press C-o C-d and wait until a documentation hint for the method call is displayed briefly in the message buffer. PerlySense will look in base classes if the method can't be found in the current class.
Put the cursor on the "method" word of an $object->method call and press C-o C-d to see the docs hint. PerlySense will look through all your "use" modules (and their base classes) for the method call and try to identify the best match.
Note! The first time each module is parsed this will take a second or two, and the very first time you run the command with lots of "use" modules it's bound to take longer than that.
Put the cursor on a module name and press C-o C-d to bring up a new buffer with the POD for that module (this is similar to the cperl-mode feature, only a) not as good, but b) it works on Windows).
Press C-o C-d with nothing under the cursor brings up a POD buffer for the current file.
C-o C-g is the "Smart go to" command. It's similar to Smart Docs, but instead of bringing the docs to you, it brings you to the definition of what's at point.
The definition can be either the sub declaration, or if the declaration can't be found (like for auto-generated getters/setters, autoloaded subs etc), the POD documentation for the sub.
Before you go anywhere the mark is set. Go back to earlier marks globally with C-x C-SPC, or locally with C-u C-SPC.
C-o m f -- Go to Class (will be changed to C-o g c) at point.
Pressing C-o C-c will bring up the Class Overview of the Class name at point (not yet implemented), or otherwise the current Class (the active Package).
Example class CatalystX::FeedMe::Controller::Feed
* Inheritance * [ Class::Accessor ] +> [ Class::Accessor::Fast ] <-----+ | [ Catalyst::AttrContainer ] ------+---------------------------+ | | | v +- [ Catalyst::Base ] --> [ Catalyst::Component ] --> [ Class::Data::Inheritable ] [ Catalyst::Controller ] [<CatalystX::FeedMe::Controller::Feed>] * Uses * [ Data::Dumper ] [ XML::Atom::Syndication::Content ] [ XML::Atom::Syndication::Feed ] [ Template::Filters ] [ XML::Atom::Syndication::Entry ] [ XML::Atom::Syndication::Link ] * NeighbourHood * [ CatalystX::FeedMe::DBIC ] [<CatalystX::FeedMe::Controller::Feed >] -none- [ CatalystX::FeedMe::Controller::FeedItem ] [ CatalystX::FeedMe::Controller::Homepage ] [ CatalystX::FeedMe::Controller::Root ] * Bookmarks * - Todo Feed.pm:83: remove duplication * API * \>mutator_name_for ->new ->path_prefix ...
The Inheritance section shows all Base classes of the Class. Inheriting from something like Catalyst is hopefully the hairiest you'll see. Classes inherit from their parents upwards in the diagram unless there is an arrow pointing elsewhere.
The Uses section shows all used modules in the Class.
The NeighbourHood section shows three columns (1: parent dir, 2: current dir, 3: subdir for the current class) with Classes located nearby (this can be bizarrely huge (and take a long time) if you browse your site_lib or similar).
The Bookmarks section shows matches for bookmark definitions you have defined in the Project config (see below).
the API section shows things that look like methods and properties of the class (sub declarations, $self method calls, $self->{hash_ref_keys}):
->method_in_this_class \>method_in_base_class (note the arrow coming from above)
When in the Class Overview buffer:
g -- Go to the file of the thing at point.
d -- Documentation for the thing at point.
c -- Class Overview for the thing at point. RET does the same.
I -- Move point to the Inheritance heading in the buffer.
U -- Move point to the Uses heading in the buffer.
H -- Move point to the NeighbourHood heading (mnemonic: 'Hood).
B -- Move point to the Bookmarks heading.
A -- Move point to the API heading.
N -- Move point to the '->new' method in the buffer (if any).
q -- Quit the Class Overview buffer.
C-o C-r -- Run the file of the current buffer using the Compilation mode.
Files are run according to the source type, which is determined by the file name (see the config file). The default for .t files is to run "prove -v", for .pm files "perl -c", etc. This can be configured per Project (see below).
The file is run from the Project root directory or from the file directory depending on the file type, and the @INC is set appropriately. You can also specify additional @INC directories in the Project config.
(Note that you can configure whatever type of run profile you like, not just Perl source files.
As a taste of what's possible, imagine that you have a test framework with .yml acceptance test data files and a corresponding yml-runner.pl script. With the x option (not implemented) you can edit the .yml file and type C-o C-r to run the acceptance test the same way as a regular test.)
If any warnings, errors or test failures are encountered, they are highlighted in the *compilation* buffer. Use C-c C-c to move from one error to the next. Or press RET on a highlighted line.
If you wish to start many runs at the same time, rename the compilation buffer with "M-x rename-buffer".
Invoke C-o C-r from within the *compilaton* buffer to re-run (M-x recompile) the file. Useful when you have skipped around the source fixing errors and the .t file isn't visible.
C-o r r -- If not even the *compilation* buffer is visible, issue Re-Run File from anywhere to bring it up and re-run.
If you run tests in a regular shell (inside Emacs or in a terminal window), this may be handy.
C-o g e -- If point is located on an error line from a syntax error, or a stack trace from the debugger or similar, go to that file+line.
If no file name can be found, prompt for a piece of text that contains the file+line spec. The kill ring or clipboard text is used as default if available (so it's easy to just copy the error line from the terminal, run this command and hit return to accept the default text).
"Flymake performs on-the-fly syntax checks of the files being edited using the external syntax check tool (usually the compiler). Highlights erroneous lines and displays associated error messages."
Flymake is included in Emacs 22 (or available from http://flymake.sourceforge.net/, put flymake.el somewhere in your load-path. [[[explain how to fix brokenness?]]] ).
PerlySense uses flymake to check syntax, Perl Critic, etc.
Three inconveniences with vanilla Flymake are fixed: no proper @INC, only .pl files, and "perl -c" warns about redefined subs for recursively used modules (which is perfectly fine Perl).
Syntax errors and warnings both use the error face.
Perl Critic violations use the warning face.
First off, flymake itself needs to be enabled. Refer to the Emacs Installation description.
This will enable Flymake for all cperl-mode buffers, causing Emacs to call perly_sense for each check.
PerlySense won't do anything at this point though. You still need to configure what should happen during a flymake.
Create a PerlySense Project directory (see below) and look in the project.yml file for instructions on how to configure Flymake activities.
Set "syntax" / "critic" to 1 to enable them.
The primary reason "syntax" is turned off by default is that it's a potential security hole; running "perl -c" on a file will not only check the syntax; BEGIN and CHECK blocks are also executed. Doing that on random code may be considered... baaad.
This way you can have Flymake enabled globally and still not run "perl -c" on everything that happens to be in a buffer.
In the Project config file there are some hints on how to customize Flymake, when it should run, etc. You can also customize it with "M-x customize-group flymake".
(Personally I find the nagging while I type very distracting, but I welcome the immediate feedback whenever I save the file. YMMV.)
Look in the mode line for hints on whether there are any errors or warnings.
C-o s n -- Go to the next Source error/warning.
Display the error in the minibuffer. If the warning is from a Perl::Critic module, copy the module name into the kill-ring, so you easily can yank it into the .perlcritic config file to disable it. (not implemented)
C-o s p -- Go to the previous Source error/warning.
C-o s s -- Display the error/warning text of the current line.
PerlySense keeps a per-user directory to store cache files, logs, etc. The .PerlySense user directory is located under the first available of these environment variables:
.PerlySense
$APPDATA $ALLUSERSPROFILE $USERPROFILE $HOME $TEMP $TMP
PerlySense has the concept of a Project root directory.
If you follow the standard directory structure for CPAN modules, the Project directory is typically the one which contains the Makefile.PL, the lib, bin, and t directory, etc.
Basically, this is where all the source lives, and where your program can go to find modules that are used. This is from where tests are run and files are found.
The fastest and most solid way for PerlySense to know which is the Project directory is to create a .PerlySenseProject directory with a config file in it. This is highly recommended for all of your own projects.
The complete project identification strategy is as follows:
First, if there is any directory upwards in the dirctory path with a .PerlySenseProject dir in it, that is the Project directory.
Second, PerlySense will try figure out from where the current file (if any) was being required/used given the contained package names or used modules.
Third, if that doesn't work, PerlySense will look for lib and t directories.
If that doesn't work, PerlySense is lost and you really do need to create an explicit Project directory by running the following command in your intended Project root directory (that would typically be the directory which has a "lib" directory in it):
perly_sense create_project
Any existing .PerlySenseProject/project.yml config file will be renamed.
The Project has a .PerlySenseProject/project.yml config file. Here you can change the name of the Project, add extra @INC directories, etc.
There is a yaml-mode for Emacs, but I haven't got it to work properly (unless an infinite loop counts as "properly" these days). The shell-script-mode is good enough.
The config file documentation is where it belongs, in the config file, so just take a look at it.
Bookmarks are regexes that may match against a single line. Each bookmark definition has a name/moniker under which the matches are grouped in the Class Overview display.
The primary point of Bookmarks is to highlight unusual things in the source. The secondary to make it easy for you go navigate to them.
This can be anything you like, but things that come to mind are:
TODO comments
FIXME/XXX/HACK comments
Things you don't want left in the code, like
Breakpoints ($DB::single = 1)
Debugging warn/print statements
Bookmarks are defined in the Project Config file (technical details are documented there).
Since Perl is so dynamic, a perfect static analysis of the source is impossible. But not unusably so. Well, hopefully. Most of the time.
Because of this PerlySense is not about exact rules, but about heuristics and a 90% solution that isn't perfect, but good-enough.
PerlySense tries to take advantage of the fact that Perl code is more than the plain source file. The source lives in a context of POD and a directory structure and common Perl idioms.
Sometimes when PerlySense can't make a decision, you're expected to chip in and tell it what you meant.
Sometimes it won't work at all.
Such is the way of dynamic languages.
If it works for you, brilliant, use it to be more productive. If not... well, there's always Java >:)
sepia - similar effort
PPI - excellent for parsing Perl
CPANXR - also uses PPI for cross referencing the CPAN
http://www.DarSerMan.com/Perl/Oasis/ - Win32 class browser/IDE. Earlier (a lot) work by me.
http://c2.com/doc/SignatureSurvey/ - The idea behind Signature Surveys. Introduced in this article about Software Archeology (http://www.pragmaticprogrammer.com/articles/mar_02_archeology.pdf).
Johan Lindström, <johanl[ÄT]DarSerMan.com>
<johanl[ÄT]DarSerMan.com>
Please report any bugs or feature requests to bug-devel-perlysense@rt.cpan.org, or through the web interface at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Devel-PerlySense. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.
bug-devel-perlysense@rt.cpan.org
Tab/space isn't supported by PPI yet, but it's supposed to be. So using Tab instead of spaces won't work properly.
PPI is kinda slow for large documents. Lots of objects being created etc.
There are certainly edge cases. Bug reports with failing tests appreciated :)
Peter Liljenberg for his elisp fu.
Copyright 2007 Johan Lindström, All Rights Reserved.
This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
Cache::Cache object, or undef if no cache is active.
Default: undef
Devel::PerlySense::Project object.
Default: A Devel::PerlySense::Project::Unknown object.
Devel::PerlySense::Home object.
Default: A newly created Home object.
Hash ref with the current config.
If there is a known Project, it reflects the Project's config, otherwise it's the default config.
Readonly. Note that the _entire_ data structure is readonly. Each time you change/add/remove a value from it, a kitten is slain. So, dude, just don't go there!
Devel::PerlySense::BookmarkConfig object.
Create new PerlySense object.
Identify a project given the $file or $dir, and set the oProject property.
If there is already a project defined, don't change it.
If no project was found, don't change oProject.
Return 1 if there is a valid project, else 0.
Die on errors.
Parse $file into a new PerlySense::Document object.
Return the new object.
Die on errors (like if the file wasn't found).
Return the pod in $file as text, or die on errors.
Die if $file doesn't exist.
Look in $file at location $row/$col and determine what is there. Depending on what's there, find the source declaration/whatever, find it and return an Devel::PerlySense::Document::Location object.
Currently supported:
$self->method, look in current file and base classes. If no sub can be found, look for POD. $object->method, look in current file and used modules. If no sub can be found, look for POD. Module::Name (bareword) Module::Name (as the only contents of a string literal)
If there's nothing at $row/col, or if the source can't be found, return undef.
Die if $file doesn't exist, or on other errors.
Look in $file at location $row/$col and determine what is there. Depending on what's there, find the documentation for it and return a Document::Location object with the following rhProperty keys set:
text - the docs text found - "method" | "module" docType - "hint" | "document" name - the name of the thing found
Same as for oLocationSmartGoTo
If there's nothing at $row/col, use the current document.
Figure out what type of source file $fileSource is, and how it should be run.
The settings in the Project's config->{run_file} is used to determine the details.
Return hash ref with (keys: "dir_run_from", "command_run", "type_source_file"), or die on errors (like if no Project could be found).
dir_run_from is an absolute file name which should be the cwd when command_run is executed.
type_source_file is something like "Test", "Module".
Do a flymake run with $fileSource according to the flymake config and output the result to STDOUT and STDERR.
Create a new PerlySense Project in $dir.
Return 1 on success, or die on errors.
Look in $file at location $row/$col and determine what class name that is.
Return the class name or "" if it's package main.
Look in $file at location $row/$col and determine what PerlySelse::Class that is.
Return the Class object or undef if it's package main.
Find the file that contains the Class $name, starting at $dirOrigin.
Return the Class object or undef if it couldn't be found.
Find the file containing the $nameModule given the $dirOrigin.
Return the absolute file name, or undef if none could be found. Die on errors.
Return a parsed PerlySense::Document, or undef if none could be found. Die on errors.
Find the file containing the $fileModuleBase given the $dirOrigin.
Return the file name relative to $dirOrigin, or undef if none could be found. Die on errors.
Find the dir containing the $fileModuleBase (relative file path) given the $dirOrigin. For all directories, also look in subdirectories in $raDirSub.
Return the absolute dir name, or undef if none could be found. Die on errors.
Find the file containing the $nameModule in @INC.
Return the $nameModule converted to a file name (i.e. with dirs and .pm extension).
Check if $fileModuleBase is located in $dir.
Return the absolute file name, or "" if not found at $dir.
Return $pod rendered as text, or die on errors.
Render the $oLocation->rhProperty->{pod} and put it in rhProperty->{text}.
Return the same (modified) $oLocation object, or undef if no rhProperty->{pod} property ended up as text (after this operation, there is content in rhProperty->{text}).
Return undef if $oLocation is undef.
Return a list with Devel::PerlySense::Document objects that support all of the methods in $raMethodRequired and possibly the methods in $raMethodNice. Look in modules in $raNameModule.
The list is sorted with the best match first.
If the document APIs have one or more base classes, look in the @ISA (depth-first, just like Perl (see perldoc perltoot)).
Warn on some failures to find the location. Die on errors.
Look in $file at location $row/$col and determine what package is there.
Return a two item array with (Package name, Devel::PerlySense::Document::Api object with the likely API of that class), or () if none was found.
Return a list with Devel::PerlySense::Document objects found under the $dir, and that return true for the grep sub $rsGrepFile and $rsGrepDocument.
If any found file couldn't be parsed, skip it silently from the list.
If the oCache isn't undef, store the $value in the cache under the total key of ($file, $file's timestamp, $key, and the PerlySense VERSION).
$value should be a scalar or reference which can be freezed.
$file must be an existing file.
Return 1 if the $value was stored, else 0. Die on errors.
If the oCache isn't undef, get the value in the cache under the total key of ($file, $file's timestamp, $key) and return it.
Return the value, or undef if the value could not be fetched. Die on errors.
If oCache is undef, return undef.
Otherwise, return the total key of ($file, $file's timestamp, $key, and the PerlySense VERSION).
1 POD Error
The following errors were encountered while parsing the POD:
Non-ASCII character seen before =encoding in 'Lindström,'. Assuming CP1252
To install Devel::PerlySense, copy and paste the appropriate command in to your terminal.
cpanm
cpanm Devel::PerlySense
CPAN shell
perl -MCPAN -e shell install Devel::PerlySense
For more information on module installation, please visit the detailed CPAN module installation guide.