D9247: Extend Scripting API to allow executong commands

[Dominik Haumann]

dhaumann created this revision.
dhaumann added reviewers: cullmann, mwolff, kfunk.
Restricted Application added projects: Kate, Frameworks.
Restricted Application added a subscriber: Frameworks.

REVISION SUMMARY
  This patch adds the following JavaScript API:
  
  object view.executeCommand(String command,
  
    String args = String(),
    Range range = Rante::invalid())
  
  The returned object indicates whether execution of the command
  was successful:
  
  - object.ok [bool] is true on success, otherwise false.
  - object.status [String] contains the result of the command (e.g. "31 replacements made", or an error message)
  
  Examples:
  
    view.executeCommand("sort"); // all, or seletcion
    view.executeCommand("sort", "", new Range(1, 0, 4, 0); // sort lines 1-4
    view.executeCommand("set-indent-width", "4");
  
  Unfortunately, we do not have any getters, e.g. "indent-width".
  This is a shortcoming of the KTextEditor::Command API. We should
  consider to change KTextEditor::Command as follows:
  
    // old:
    virtual bool exec(KTextEditor::View *view, const QString &cmd, QString &msg,
                      const KTextEditor::Range &range = KTextEditor::Range::invalid()) = 0;
    
    // new:
    virtual QVariant exec(const QVariant & args,
                      const KTextEditor::Range &range = KTextEditor::Range::invalid(),
                      KTextEditor::View * view = nullptr) = 0;
  
  or similar. This way, Commands would be much more usabe in general
  (and also better chainable, which could result in: command 1 | command 2 | ...).
  
  What is missing in this patch:
  
  - infinite recursion is very easily possible
  - guard against some special commands: reload-scripts, anything else?
  - should the API be more smart? E.g. we could support both:
    - view.executeCommand("sort", "", new Range(1,0, 3, 0))
    - view.executeCommand("sort", new Range(1,0, 3, 0)) -> we'd have to lookup the type of the passed QJSValues (would work)...
  - Maybe only return a bool instead of object with "ok" and "status" ? -> The status could be helpful as a return value, but to be honest, this is a hack (see shortcoming of KTextEditor::Command above). We could also add KTextEditor::CommandExtension, provide an improved interface, and then do this properly here... Comments?

TEST PLAN
  make test, manual testing

REPOSITORY
  R39 KTextEditor

BRANCH
  CommandScript (branched from master)

REVISION DETAIL
  https://phabricator.kde.org/D9247

AFFECTED FILES
  src/script/katescriptview.cpp
  src/script/katescriptview.h

To: dhaumann, cullmann, mwolff, kfunk
Cc: #frameworks, kevinapavew, ngraham, demsking, cullmann, sars, dhaumann
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20171208/e0fd95a3/attachment-0001.html>