(*:maxLineLen=78:*)
theory JEdit
imports Base
begin
chapter ‹Introduction
›
section ‹Concepts
and terminology
›
text ‹
Isabelle/jEdit
is a Prover IDE that integrates
🚫‹parallel
proof checking
›
🍋‹"Wenzel:2009" and "Wenzel:2013:ITP"› with 🚫‹asynchronous user
interaction
› 🍋‹"Wenzel:2010" and "Wenzel:2012:UITP-EPTCS" and
"Wenzel:2014:ITP-PIDE" and "Wenzel:2014:UITP"›, based on a document-oriented
approach
to 🚫‹continuous
proof processing
› 🍋‹"Wenzel:2011:CICM" and
"Wenzel:2012" and "Wenzel:2018:FIDE" and "Wenzel:2019:MKM"›. Many concepts
and system components are fit together
in order
to make this work. The main
building blocks are as follows.
🚫[Isabelle/ML]
is the implementation
and extension language of Isabelle,
see
also 🍋‹"isabelle-implementation"›. It
is integrated into the
logical
context of Isabelle/Isar
and allows
to manipulate logical entities
directly. Arbitrary add-on tools may be implemented
for object-logics such
as Isabelle/HOL.
🚫[Isabelle/Scala]
is the system programming language of Isabelle. It
extends the pure logical environment of Isabelle/ML towards the outer
world of graphical user interfaces,
text editors, IDE frameworks, web
services, SSH servers, SQL databases etc. Both Scala
and ML provide
library modules
to support formatted
text with formal markup,
and to
encode/decode algebraic datatypes. Scala communicates
with ML via
asynchronous protocol commands;
from the ML perspective this
is wrapped up
as synchronous
function call (RPC).
🚫[PIDE]
is a general framework
for Prover IDEs based on Isabelle/Scala. It
is built around a concept of parallel
and asynchronous document
processing, which
is supported natively
by the parallel
proof engine that
is implemented
in Isabelle/ML. The traditional prover command loop
is
given up; instead there
is direct support
for editing of source
text,
with
rich formal markup
for GUI rendering.
🚫[jEdit]
is a sophisticated
text editor
🚫‹🚫‹http://www.jedit.org››
implemented
in Java
🚫‹🚫‹https://openjdk.java.net››. The editor
is easily
extensible
by plugins written
in any language that works on the JVM.
In
the
context of Isabelle this
is usually
Scala
🚫‹🚫‹https://www.scala-lang.org››.
🚫[Isabelle/jEdit]
is the main application of the PIDE framework
and the
default user-interface
for Isabelle. It targets both beginners
and
experts. Technically, Isabelle/jEdit consists of the original jEdit code
base
with minimal patches
and a special plugin
for Isabelle. This
is
integrated as a desktop application
for the main operating system
families: Linux, Windows, macOS.
End-users of Isabelle download
and run a standalone application that exposes
jEdit as a
text editor on the surface.
Thus there
is occasionally a tendency
to apply the name ``jEdit
'' to any of the Isabelle Prover IDE aspects,
without proper differentiation. When discussing these PIDE building blocks
in public forums, mailing lists, or even scientific publications, it
is
particularly important
to distinguish Isabelle/ML versus Standard ML,
Isabelle/Scala versus Scala, Isabelle/jEdit versus jEdit.
›
section ‹The Isabelle/jEdit Prover IDE
›
text ‹
\begin{figure}[!htb]
\begin{center}
\includegraphics[width=
\textwidth]{isabelle-jedit}
\end{center}
\caption{The Isabelle/jEdit Prover IDE}
\label{fig:isabelle-jedit}
\end{figure}
Isabelle/jEdit (
\figref{fig:isabelle-jedit}) consists of some plugins
for
the jEdit
text editor, together
with patches on the underlying code base.
The overall look-and-feel of jEdit
is changed significantly,
and the
different name of Isabelle/jEdit
is justified even
from the surface.
The main plugin
is called ``Isabelle
'' and has its own menu
🚫‹Plugins~/
Isabelle
› with access
to several
actions and add-on panels (see
also
\secref{sec:dockables}), as well as
🚫‹Plugins~/ Plugin Options~/ Isabelle
›
(see
also \secref{sec:options}).
The plugin options allow
to specify a logic session name, but the same
selector
is also accessible
in the
🚫‹Theories
› panel
(
\secref{sec:theories}). After startup of the Isabelle plugin, the selected
logic session image
is provided automatically
by the Isabelle build tool
🍋‹"isabelle-system"›:
if it
is absent or outdated wrt.
\ its sources,
the build process updates it within the running
text editor. Prover IDE
functionality
is only activated after successful
termination of the build
process. A failure may require changing some options
and restart of the
Isabelle plugin or application. Changing the logic session requires a
restart of the whole application
to take effect.
🚫 The main job of the Prover IDE
is to manage sources
and their changes,
taking the logical
structure as a formal document into account (see
also
\secref{sec:document-model}). The editor
and the prover are connected
asynchronously without locking. The prover
is free
to organize the checking
of the formal
text in parallel on multiple cores,
and provides feedback via
markup, which
is rendered
in the editor via colors, boxes, squiggly
underlines, hyperlinks, popup windows, icons, clickable
output etc.
Using the mouse together
with the modifier key
🍋‹CONTROL
› (Linux, Windows)
or
🍋‹COMMAND
› (macOS) exposes formal content via tooltips
and/or
hyperlinks (see
also \secref{sec:tooltips-hyperlinks}).
Output (
in popups
etc.) may be explored recursively,
using the same techniques as
in the
editor source buffer.
Thus the Prover IDE gives an impression of direct access
to formal content
of the prover within the editor, but
in reality only certain aspects are
exposed, according
to the possibilities of the prover
and its add-on tools.
›
subsection ‹Documentation
›
text ‹
The
🚫‹Documentation
› panel of Isabelle/jEdit provides access
to some example
theory files
and the standard Isabelle documentation. PDF files are opened
by regular desktop operations of the underlying platform. The
section
``Original jEdit Documentation
'' contains the original
🚫‹User
's Guide\ of
this sophisticated
text editor. The same
is accessible via the
🍋‹Help› menu
or
🍋‹F1
› keyboard shortcut,
using the built-in HTML viewer of Java/Swing.
The latter
also includes 🚫‹Frequently Asked Questions
› and documentation of
individual plugins.
Most of the information about jEdit
is relevant
for Isabelle/jEdit as well,
but users need
to keep
in mind that defaults sometimes differ,
and the
official jEdit documentation does not know about the Isabelle plugin
with
its support
for continuous checking of formal source
text: jEdit
is a plain
text editor, but Isabelle/jEdit
is a Prover IDE.
›
subsection ‹Plugins
›
text ‹
The
🚫‹Plugin Manager
› of jEdit allows
to augment editor functionality
by JVM
modules (jars) that are provided
by the central plugin repository, which
is
accessible via various mirror sites.
Connecting
to the plugin server-infrastructure of the jEdit project allows
to update bundled plugins or
to add further functionality. This needs
to be
done with the usual care
for such an
open bazaar of contributions, many of
them unmaintained. Arbitrary combinations of add-on features are apt
to
cause problems. It
is advisable
to start
with the default configuration of
Isabelle/jEdit
and develop a sense how it
is meant
to work, before loading
other plugins.
🚫
The
🚫‹Isabelle
› plugin
is responsible
for the main Prover IDE functionality
of Isabelle/jEdit: it manages the prover session
in the background. A few
additional plugins are bundled
with Isabelle/jEdit
for convenience or out of
necessity, notably
🚫‹Console
› with its
🚫‹Scala
› sub-plugin
(
\secref{sec:scala-console})
and 🚫‹SideKick
› with some Isabelle-specific
parsers
for document tree
structure (
\secref{sec:sidekick}). Other plugins
(e.g.
\ 🚫‹Console
›,
🚫‹ErrorList
›,
🚫‹SideKick
›) are included
to saturate the
dependencies of bundled plugins, but
have no particular
use in
Isabelle/jEdit.
›
subsection ‹Options
\label{sec:options}
›
text ‹
Both jEdit
and Isabelle
have distinctive management of persistent options.
Regular jEdit options are accessible via the dialogs
🚫‹Utilities~/
Global
Options
› or
🚫‹Plugins~/ Plugin Options
›,
with a second chance
to flip the
two within the central options dialog. Changes are stored
in
🍋‹$JEDIT_SETTINGS/properties
› and 🍋‹$JEDIT_SETTINGS/keymaps
›.
Isabelle system options are managed
by Isabelle/Scala
and changes are stored
in 🍋‹$ISABELLE_HOME_USER/etc/preferences
›, independently of
other jEdit properties. See
also 🍋‹"isabelle-system"›, especially the
coverage of sessions
and command-line tools like @{tool build} or @{tool
options}.
Those Isabelle options that are declared as
🍋‹public
› are configurable
in
Isabelle/jEdit via
🚫‹Plugin Options~/ Isabelle~/ General
›.
Moreover, there
are various options
for rendering document content, which are configurable
via
🚫‹Plugin Options~/ Isabelle~/ Rendering
›.
Thus 🚫‹Plugin Options~/
Isabelle
› in jEdit provides a view on a subset of Isabelle system options.
Note that some of these options affect general parameters that are relevant
outside Isabelle/jEdit as well, e.g.
\ @{system_option threads} or
@{system_option parallel_proofs}
for the Isabelle build tool
🍋‹"isabelle-system"›, but it
is possible
to use the settings variable
@{setting ISABELLE_BUILD_OPTIONS}
to change defaults
for batch builds on the
command-line, without affecting the Prover IDE.
The jEdit action @{action_def isabelle.options} opens the options dialog
for
the Isabelle plugin; it can be mapped
to editor GUI elements as usual.
🚫
Options are usually loaded on startup
and saved on shutdown of
Isabelle/jEdit. Editing the generated
🍋‹$JEDIT_SETTINGS/properties
›
or
🍋‹$ISABELLE_HOME_USER/etc/preferences
› manually while the
application
is running may cause lost updates!
›
subsection ‹Keymaps
›
text ‹
Keyboard shortcuts are managed as a separate concept of
🚫‹keymap
› that
is
configurable via
🚫‹Global Options~/ Shortcuts
›. The
🍋‹imported
› keymap
is
derived
from the initial environment of properties that was available at the
first start of the editor; afterwards the keymap
file takes precedence
and
is no longer affected
by change of default properties.
Users may modify their keymap later, but this can lead
to conflicts
with
🍋‹shortcut
› properties
in 🍋‹$JEDIT_HOME/properties/jEdit.props
›.
The action @{action_def
"isabelle.keymap-merge"} helps
to resolve pending
Isabelle keymap changes wrt. the current jEdit keymap; non-conflicting
changes are applied implicitly. This action
is automatically invoked on
Isabelle/jEdit startup.
›
section ‹Command-line invocation
\label{sec:command-line}
›
text ‹
Isabelle/jEdit
is normally invoked as a single-instance desktop application,
based on platform-specific executables
for Linux, Windows, macOS.
It
is also possible
to invoke the Prover IDE on the command-line,
with some
extra options
and environment settings. The command-line usage of @{tool_def
jedit}
is as follows:
@{verbatim [display]
‹Usage: isabelle jedit [OPTIONS] [FILES ...]
Options are:
-A NAME ancestor session
for option -R (default: parent)
-D NAME=X set JVM system property
-J OPTION add JVM runtime option
(default $JEDIT_JAVA_SYSTEM_OPTIONS $JEDIT_JAVA_OPTIONS)
-R NAME build image
with requirements
from other sessions
-b build only
-d DIR include session directory
-f fresh build
-i NAME include session
in name-space of theories
-j OPTION add jEdit runtime option
(default $JEDIT_OPTIONS)
-l NAME logic session name
-m MODE add print mode
for output
-n no build of session image on startup
-o OPTION override Isabelle system OPTION (via NAME=VAL or NAME)
-p CMD command prefix
for ML process (e.g. NUMA policy)
-s system build mode
for session image (system_heaps=true)
-u user build mode
for session image (system_heaps=false)
Start jEdit
with Isabelle plugin
setup and open FILES
(default
"$USER_HOME/Scratch.thy" or
":" for empty buffer).
›}
The
🍋‹-l
› option specifies the session name of the logic image
to be used
for proof processing. Additional session root directories may be included
via option
🍋‹-d
› to augment the session name space (see
also 🍋‹"isabelle-system"›).
By defa
ult, the specified image is checked and built on
demand, but option 🍋‹-n› bypasses the implicit build process for the
selected session image. Options 🍋‹-s› and 🍋‹-u› override the default system
option @{system_option system_heaps}: this determines where to store the
session image of @{tool build}.
The 🍋‹-R› option builds an auxiliary logic image with all theories from
other sessions that are not already present in its parent; it also opens the
session 🍋‹ROOT› entry in the editor to facilitate editing of the main
session. The 🍋‹-A› option specifies and alternative ancestor session for
option 🍋‹-R›: this allows to restructure the hierarchy of session images on
the spot. Options 🍋‹-R› and 🍋‹-l› are mutually exclusive.
The 🍋‹-i› option includes additional sessions into the name-space of
theories: multiple occurrences are possible.
The 🍋‹-m› option specifies additional print modes for the prover process.
Note that the system option @{system_option_ref jedit_print_mode} allows to
do the same persistently (e.g.\ via the 🚫‹Plugin Options› dialog of
Isabelle/jEdit), without requiring command-line invocation.
The 🍋‹-J› and 🍋‹-j› options pass additional low-level options to the JVM or
jEdit, respectively. The defaults are provided by the Isabelle settings
environment 🍋‹"isabelle-system"›, but note that these only work for the
command-line tool described here, and not the desktop application.
The 🍋‹-D› option allows to define JVM system properties; this is passed
directly to the underlying 🍋‹java› process.
The 🍋‹-b› and 🍋‹-f› options control the self-build mechanism of
Isabelle/Scala. This is only relevant for building from sources, the
official Isabelle release already includes a pre-built version of
everything required for Isabelle/jEdit.
The 🍋‹-o› option is analogous to @{tool build} 🍋‹"isabelle-system"›,
but it takes persistent preferences into account (\secref{sec:options}).
When options are loaded, command-line options take precedence. When options
are saved, command-line options are ignored (despite subsequent changes),
but original preferences take precedence (including subsequent changes).
🚫
It is also possible to connect to an already running Isabelle/jEdit process
via @{tool_def jedit_client}:
@{verbatim [display]
‹Usage: isabelle jedit_client [OPTIONS] [FILES ...]
Options are:
-c only check presence of server
-n only report server name
-s NAME server name (default "Isabelle")
Connect to already running Isabelle/jEdit instance and open FILES›}
The 🍋‹-c› option merely checks the presence of the server, producing a
process return-code.
The 🍋‹-n› option reports the server name, and the 🍋‹-s› option provides a
different server name. The default server name is the official distribution
name (e.g.\ 🍋‹Isabelle2025-1›). Thus @{tool jedit_client} can connect to the
Isabelle desktop application without further options.
The 🍋‹-p› option allows to override the implicit default of the system
option @{system_option_ref process_policy} for ML processes started by
the Prover IDE, e.g. to control CPU affinity on multiprocessor systems.
The JVM system property 🍋‹isabelle.jedit_server› provides a different server
name, e.g.\ use 🍋‹isabelle jedit -Disabelle.jedit_server=›‹name› and
🍋‹isabelle jedit_client -s›~‹name› to connect later on.
›
section ‹GUI rendering›
text ‹
Isabelle/jEdit is a classic Java/AWT/Swing application: its GUI rendering
usually works well, but there are technical side-conditions on the Java
window system and graphics engine. When researching problems and solutions
on the Web, it often helps to include other well-known Swing applications,
notably IntelliJ IDEA and Netbeans.
›
subsection ‹Portable and scalable look-and-feel›
text ‹
In the past, 🚫‹system look-and-feels› tried hard to imitate native GUI
elements on specific platforms (Windows, macOS/Aqua, Linux/GTK+), but many
technical problems have accumulated in recent years (e.g.\ see
\secref{sec:problems}).
Already since 2021, we are de-facto back to 🚫‹portable look-and-feels›,
which also happen to be 🚫‹scalable› on high-resolution displays:
🚫 🍋‹FlatLaf Light› (or 🍋‹FlatLaf macOS Light›) is enabled by default. It
generally looks good on all platforms, and works smoothly with
high-resolution displays.
🚫 🍋‹FlatLaf Dark› (or 🍋‹FlatLaf macOS Dark›) is a notable alternative. It
indicates that 🚫‹dark mode› should be used for rendering in
Isabelle/jEdit, via jEdit options with suffix ``🍋‹.dark›'' and Isabelle
options with suffix ``🍋‹_dark›''. The panels for 🚫‹Global Options› and
🚫‹Plugin Options / Isabelle / Rendering› operate on options according to
the current Swing look-and-feel, e.g. on 🍋‹view.fgColor.dark› in dark mode
vs. 🍋‹view.fgColor› in non-dark mode.
🚫 🍋‹Metal› still works, although it is stylistically outdated. It might
require manual adjustments of font sizes for high-resolution displays (see
\secref{sec:fonts}).
Most other look-and-feels are better ignored: they look rather bad, or cause
genuine problems with GUI interaction.
Changing the look-and-feel in 🚫‹Global Options~/ Appearance› updates the GUI
only partially: a full restart of Isabelle/jEdit is required to see the true
effect.
›
subsection ‹Adjusting fonts \label{sec:fonts}›
text ‹
The preferred font collection for Isabelle/jEdit is 🍋‹Isabelle DejaVu›: it
is used by default for the main text area and various GUI elements. The
default font sizes attempt to deliver a usable application for common
display types, such as ``Full HD'' at $1920 \times 1080$ and ``Ultra HD'' at
$3840 \times 2160$.
🚫 Isabelle/jEdit provides various options to adjust font sizes in particular
GUI elements. Here is a summary of all relevant font properties:
🚫 🚫‹Global Options / Text Area / Text font›: the main text area font,
which is also used as reference point for various derived font sizes,
e.g.\ the 🚫‹Output› (\secref{sec:output}) and 🚫‹State›
(\secref{sec:state-output}) panels.
🚫 🚫‹Global Options / Gutter / Gutter font›: the font for the gutter area
left of the main text area, e.g.\ relevant for display of line numbers
(disabled by default).
🚫 🚫‹Global Options / Appearance / Button, menu and label font› as well as
🚫‹List and text field font›: this specifies the primary and secondary font
for the 🚫‹Metal› look-and-feel.
🚫 🚫‹Plugin Options / Isabelle / General / Reset Font Size›: the main text
area font size for action @{action_ref "isabelle.reset-font-size"}, e.g.java.lang.NullPointerException
relevant for quick scaling like in common web browsers.
🚫 🚫‹Plugin Options / Console / General / Font›: the console window font,
e.g.\ relevant for Isabelle/Scala command-line.
›
chapter ‹Augmented jEdit functionality›
section ‹Dockable windows \label{sec:dockables}›
text ‹
In jEdit terminology, a 🚫‹view› is an editor window with one or more 🚫‹text
areas› that show the content of one or more 🚫‹buffers›. A regular view may
be surrounded by 🚫‹dockable windows› that show additional information in
arbitrary format, not just text; a 🚫‹plain view› does not allow dockables.
The 🚫‹dockable window manager› of jEdit organizes these dockable windows,
either as 🚫‹floating› windows, or 🚫‹docked› panels within one of the four
margins of the view. There may be any number of floating instances of some
dockable window, but at most one docked instance; jEdit actions that address
🚫‹the› dockable window of a particular kind refer to the unique docked
instance.
Dockables are used routinely in jEdit for important functionality like
🚫‹HyperSearch Results› or the 🚫‹File System Browser›. Plugins often provide
a central dockable to access their main functionality, which may be opened
by the user on demand. The Isabelle/jEdit plugin takes this approach to the
extreme: its plugin menu provides the entry-points to many panels that are
managed as dockable windows. Some important panels are docked by default,
e.g.\ 🚫‹Documentation›, 🚫‹State›, 🚫‹Theories› 🚫‹Output›, 🚫‹Query›. The user
can change this arrangement easily and persistently.
Compared to plain jEdit, dockable window management in Isabelle/jEdit is
slightly augmented according to the the following principles:
🚫 Floating windows are dependent on the main window as 🚫‹dialog› in
the sense of Java/AWT/Swing. Dialog windows always stay on top of the view,
which is particularly important in full-screen mode. The desktop environment
of the underlying platform may impose further policies on such dependent
dialogs, in contrast to fully independent windows, e.g.\ some window
management functions may be missing.
🚫 Keyboard focus of the main view vs.\ a dockable window is carefully
managed according to the intended semantics, as a panel mainly for output or
input. For example, activating the 🚫‹Output› (\secref{sec:output}) or State
(\secref{sec:state-output}) panel via the dockable window manager returns
keyboard focus to the main text area, but for 🚫‹Query› (\secref{sec:query})
or 🚫‹Sledgehammer› \secref{sec:sledgehammer} the focus is given to the main
input field of that panel.
🚫 Panels that provide their own text area for output have an additional
dockable menu item 🚫‹Detach›. This produces an independent copy of the
current output as a floating 🚫‹Info› window, which displays that content
independently of ongoing changes of the PIDE document-model. Note that
Isabelle/jEdit popup windows (\secref{sec:tooltips-hyperlinks}) provide a
similar 🚫‹Detach› operation as an icon.
›
section ‹Isabelle symbols \label{sec:symbols}›
text ‹
Isabelle sources consist of 🚫‹symbols› that extend plain ASCII to allow
infinitely many mathematical symbols within the formal sources. This works
without depending on particular encodings and varying Unicode
standards.🚫‹Raw Unicode characters within formal sources compromise
portability and reliability in the face of changing interpretation of
special features of Unicode, such as Combining Characters or Bi-directional
Text.› For further explanations, see 🍋‹"Wenzel:2011:CICM"›.
For the prover back-end, formal text consists of ASCII characters that are
grouped according to some simple rules, e.g.\ as plain ``🍋‹a›'' or symbolic
``🍋‹α›''. For the editor front-end, a certain subset of symbols is rendered
physically via Unicode glyphs, to show ``🍋‹α›'' as ``‹α›'', for example.
This symbol interpretation is specified by the Isabelle system distribution
in 🍋‹$ISABELLE_HOME/etc/symbols› and may be augmented by the user in
🍋‹$ISABELLE_HOME_USER/etc/symbols›.
The appendix of 🍋‹"isabelle-isar-ref"› gives an overview of the
standard interpretation of finitely many symbols from the infinite
collection. Uninterpreted symbols are displayed literally, e.g.java.lang.NullPointerException
``🍋‹🚫›''. Overlap of Unicode characters used in symbol
interpretation with informal ones (which might appear e.g.\ in comments)
needs to be avoided. Raw Unicode characters within prover source files
should be restricted to informal parts, e.g.\ to write text in non-latin
alphabets in comments (excluding actual Greek).
›
paragraph ‹Encoding.›
text ‹Technically, the Unicode interpretation of Isabelle symbols is an
🚫‹encoding› called 🍋‹UTF-8-Isabelle› in jEdit (🚫‹not› in the underlying
JVM). It is provided by the Isabelle Base plugin and enabled by default for
all source files in Isabelle/jEdit.
Sometimes such defaults are reset accidentally, or malformed UTF-8 sequences
in the text force jEdit to fall back on a different encoding like
🍋‹ISO-8859-15›. In that case, verbatim ``🍋‹α›'' will be shown in the text
buffer instead of its Unicode rendering ``‹α›''. The jEdit menu operation
🚫‹File~/ Reload with Encoding~/ UTF-8-Isabelle› helps to resolve such
problems (after repairing malformed parts of the text).
If the loaded text already contains Unicode sequences that are in conflict
with the Isabelle symbol encoding, the fallback-encoding UTF-8 is used and
Isabelle symbols remain in literal 🍋‹🚫› form. The jEdit menu
operation 🚫‹Utilities~/ Buffer Options~/ Character encoding› allows to
enforce 🍋‹UTF-8-Isabelle›, but this will also change original Unicode text
into Isabelle symbols when saving the file!
›
paragraph ‹Font.›
text ‹Correct rendering via Unicode requires a font that contains glyphs for
the corresponding codepoints. There are also various unusual symbols with
particular purpose in Isabelle, e.g.\ control symbols and very long arrows.
Isabelle/jEdit prefers its own font collection 🍋‹Isabelle DejaVu›, with
families 🍋‹Serif› (default for help texts), 🍋‹Sans› (default for GUI
elements), 🍋‹Mono Sans› (default for text area). This ensures that all
standard Isabelle symbols are shown on the screen (or printer) as expected.
Note that a Java/AWT/Swing application can load additional fonts only if
they are not installed on the operating system already! Outdated versions of
Isabelle fonts that happen to be provided by the operating system prevent
Isabelle/jEdit to use its bundled version. This could lead to missing glyphs
(black rectangles), when the system version of a font is older than the
application version. This problem can be avoided by refraining to
``install'' a copy of the Isabelle fonts in the first place, although it
might be tempting to use the same font in other applications.
HTML pages generated by Isabelle refer to the bundled Isabelle fonts as a
server-side resource. Thus a web-browser can use that without requiring a
locally installed copy.
›
paragraph ‹Input methods.›
text ‹In principle, Isabelle/jEdit could delegate the problem to produce
Isabelle symbols in their Unicode rendering to the underlying operating
system and its 🚫‹input methods›. Regular jEdit also provides various ways to
work with 🚫‹abbreviations› to produce certain non-ASCII characters. Since
none of these standard input methods work satisfactorily for the
mathematical characters required for Isabelle, various specific
Isabelle/jEdit mechanisms are provided.
This is a summary for practically relevant input methods for Isabelle
symbols.
🚫 The 🚫‹Symbols› panel: some GUI buttons allow to insert certain symbols in
the text buffer. There are also tooltips to reveal the official Isabelle
representation with some additional information about 🚫‹symbol
abbreviations› (see below).
🚫 Copy/paste from decoded source files: text that is already rendered as
Unicode can be re-used for other text. This also works between different
applications, e.g.\ Isabelle/jEdit and some web browser or mail client, as
long as the same Unicode interpretation of Isabelle symbols is used.
🚫 Copy/paste from prover output within Isabelle/jEdit. The same principles
as for text buffers apply, but note that 🚫‹copy› in secondary Isabelle/jEdit
windows works via the keyboard shortcuts 🍋‹C+c› or 🍋‹C+INSERT›, while jEdit
menu actions always refer to the primary text area!
🚫 Completion provided by the Isabelle plugin (see \secref{sec:completion}).
Isabelle symbols have a canonical name and optional abbreviations. This can
be used with the text completion mechanism of Isabelle/jEdit, to replace a
prefix of the actual symbol like 🍋‹λ›, or its name preceded by backslash
🍋‹\lambda›, or its ASCII abbreviation 🍋‹%› by the Unicode rendering.
The following table is an extract of the information provided by the
standard 🍋‹$ISABELLE_HOME/etc/symbols› file:
🚫
\begin{tabular}{lll}
🚫‹symbol› & 🚫‹name with backslash› & 🚫‹abbreviation› \\\hline
‹λ› & 🍋‹\lambda› & 🍋‹%› \\
‹==>› & 🍋‹\Rightarrow› & 🍋‹=>› \\
‹==>› & 🍋‹\Longrightarrow› & 🍋‹==>› \\[0.5ex]
‹∧› & 🍋‹\And› & 🍋‹!!› \\
‹≡› & 🍋‹\equiv› & 🍋‹==› \\[0.5ex]
‹∀› & 🍋‹\forall› & 🍋‹!› \\
‹∃› & 🍋‹\exists› & 🍋‹?› \\
‹⟶› & 🍋‹\longrightarrow› & 🍋‹-->› \\
‹∧› & 🍋‹\and› & 🍋‹&› \\
‹∨› & 🍋‹\or› & 🍋‹|› \\
‹¬› & 🍋‹\not› & 🍋‹~› \\
‹≠› & 🍋‹\noteq› & 🍋‹~=› \\
‹∈› & 🍋‹\in› & 🍋‹:› \\
‹∉› & 🍋‹\notin› & 🍋‹~:› \\
\end{tabular}
🚫
Note that the above abbreviations refer to the input method. The logical
notation provides ASCII alternatives that often coincide, but sometimes
deviate. This occasionally causes user confusion with old-fashioned Isabelle
source that use ASCII replacement notation like 🍋‹!› or 🍋‹ALL› directly in
the text.
On the other hand, coincidence of symbol abbreviations with ASCII
replacement syntax syntax helps to update old theory sources via explicit
completion (see also 🍋‹C+b› explained in \secref{sec:completion}).
›
paragraph ‹Control symbols.›
text ‹There are some special control symbols to modify the display style of a
single symbol (without nesting). Control symbols may be applied to a region
of selected text, either using the 🚫‹Symbols› panel or keyboard shortcuts or
jEdit actions. These editor operations produce a separate control symbol for
each symbol in the text, in order to make the whole text appear in a certain
style.
🚫
\begin{tabular}{llll}
🚫‹style› & 🚫‹symbol› & 🚫‹shortcut› & 🚫‹action› \\\hline
superscript & 🍋‹🚫› & 🍋‹C+e UP› & @{action_ref "isabelle.control-sup"} \\
subscript & 🍋‹🚫› & 🍋‹C+e DOWN› & @{action_ref "isabelle.control-sub"} \\
bold face & 🍋‹🚫› & 🍋‹C+e RIGHT› & @{action_ref "isabelle.control-bold"} \\
emphasized & 🍋‹🚫› & 🍋‹C+e LEFT› & @{action_ref "isabelle.control-emph"} \\
reset & & 🍋‹C+e BACK_SPACE› & @{action_ref "isabelle.control-reset"} \\
\end{tabular}
🚫
To produce a single control symbol, it is also possible to complete on
🍋‹\sup›, 🍋‹\sub›, 🍋‹\bold›, 🍋‹\emph› as for regular symbols.
The emphasized style only takes effect in document output (when used with a
cartouche), but not in the editor.
›
section ‹Scala console \label{sec:scala-console}›
text ‹
The 🚫‹Console› plugin manages various shells (command interpreters), e.g.java.lang.NullPointerException
🚫‹BeanShell›, which is the official jEdit scripting language, and the
cross-platform 🚫‹System› shell. Thus the console provides similar
functionality than the Emacs buffers 🍋‹*scratch*› and 🍋‹*shell*›.
Isabelle/jEdit extends the repertoire of the console by 🚫‹Scala›, which is
the default. This is the regular Scala toplevel loop running inside the same
JVM process as Isabelle/jEdit itself. So the Scala command interpreter has
access to the JVM name space and state of the running Prover IDE
application. The default environment imports the full content of packages
🍋‹isabelle› and 🍋‹isabelle.jedit›.
For example, 🍋‹PIDE› refers to the Isabelle/jEdit plugin object, and 🍋‹view›
to the current editor view of jEdit. The Scala expression
🍋‹PIDE.snapshot(view)› makes a PIDE document snapshot of the current buffer
within the current editor view: it allows to retrieve document markup in a
timeless~/ stateless manner, while the prover continues its processing.
This helps to explore Isabelle/Scala functionality interactively. Some care
is required to avoid interference with the internals of the running
application.
›
section ‹Physical and logical files \label{sec:files}›
text ‹
File specifications in jEdit follow various formats and conventions
according to 🚫‹Virtual File Systems›, which may be also provided by plugins.
This allows to access remote files via the 🍋‹https:› protocol prefix, for
example. Isabelle/jEdit attempts to work with the file-system model of jEdit
as far as possible. In particular, theory sources are passed from the editor
to the prover, without indirection via the file-system. Thus files don't
need to be saved: the editor buffer content is used directly.
›
subsection ‹Local files and environment variables \label{sec:local-files}›
text ‹
Local files (without URL notation) are particularly important. The file path
notation is that of the Java Virtual Machine on the underlying platform. On
Windows the preferred form uses backslashes, but happens to accept forward
slashes like Unix/POSIX as well. Further differences arise due to Windows
drive letters and network shares: thus relative paths (with forward slashes)
are portable, but absolute paths are not.
File paths in Java are distinct from Isabelle; the latter uses POSIX
notation with forward slashes on 🚫‹all› platforms. Isabelle/ML on Windows
uses Unix-style path notation, with drive letters according to Cygwin (e.g.java.lang.NullPointerException
🍋‹/cygdrive/c›). Environment variables from the Isabelle process may be used
freely, e.g.\ 🍋‹$ISABELLE_HOME/etc/symbols› or 🍋‹$POLYML_HOME/README›.
There are special shortcuts: 🍋‹~› for 🍋‹$USER_HOME› and 🍋‹~~› for
🍋‹$ISABELLE_HOME›.
🚫 Since jEdit happens to support environment variables within file
specifications as well, it is natural to use similar notation within the
editor, e.g.\ in the file-browser. This does not work in full generality,
though, due to the bias of jEdit towards platform-specific notation and of
Isabelle towards POSIX. Moreover, the Isabelle settings environment is not
accessible when starting Isabelle/jEdit via the desktop application wrapper,
in contrast to @{tool jedit} run from the command line
(\secref{sec:command-line}).
Isabelle/jEdit imitates important system settings within the Java process
environment, in order to allow easy access to these important places from
the editor: 🍋‹$ISABELLE_HOME›, 🍋‹$ISABELLE_HOME_USER›, 🍋‹$JEDIT_HOME›,
🍋‹$JEDIT_SETTINGS›. The file browser of jEdit also includes 🚫‹Favorites› for
these locations.
🚫 Path specifications in prover input or output usually include formal
markup that turns it into a hyperlink (see also
\secref{sec:tooltips-hyperlinks}). This allows to open the corresponding
file in the text editor, independently of the path notation. If the path
refers to a directory, it is opened in the jEdit file browser.
Formally checked paths in prover input are subject to completion
(\secref{sec:completion}): partial specifications are resolved via directory
content and possible completions are offered in a popup.
›
subsection ‹PIDE resources via virtual file-systems›
text ‹
The jEdit file browser is docked by default. It provides immediate access to
the local file-system, as well as important Isabelle resources via the
🚫‹Favorites› menu. Environment variables like 🍋‹$ISABELLE_HOME› are
discussed in \secref{sec:local-files}. Virtual file-systems are more
special: the idea is to present structured information like a directory
tree. The following URLs are offered in the 🚫‹Favorites› menu, or by
corresponding jEdit actions.
🚫 URL 🍋‹isabelle-export:› or action @{action_def
"isabelle-export-browser"} shows a toplevel directory with theory names:
each may provide its own tree structure of session exports. Exports are
like a logical file-system for the current prover session, maintained as
Isabelle/Scala data structures and written to the session database
eventually. The 🍋‹isabelle-export:› URL exposes the current content
according to a snapshot of the document model. The file browser is 🚫‹not›
updated continuously when the PIDE document changes: the reload operation
needs to be used explicitly. A notable example for exports is the command
@{command_ref export_code} 🍋‹"isabelle-isar-ref"› (e.g.\ see
🍋‹$ISABELLE_HOME/src/HOL/ex/Code_Lazy_Demo.thy›).
🚫 URL 🍋‹isabelle-session:› or action @{action_def
"isabelle-session-browser"} show the structure of session chapters and
sessions within them. What looks like a file-entry is actually a reference
to the session definition in its corresponding 🍋‹ROOT› file. The latter is
subject to Prover IDE markup, so the session theories and other files may
be browsed quickly by following hyperlinks in the text.
›
section ‹Indentation›
text ‹
Isabelle/jEdit augments the existing indentation facilities of jEdit to take
the structure of theory and proof texts into account. There is also special
support for unstructured proof scripts (🚫‹apply› etc.).
🚫[Syntactic indentation] follows the outer syntax of Isabelle/Isar.
Action @{action "indent-lines"} (shortcut 🍋‹C+i›) indents the current line
according to command keywords and some command substructure: this
approximation may need further manual tuning.
Action @{action "isabelle.newline"} (shortcut 🍋‹ENTER›) indents the old
and the new line according to command keywords only: leading to precise
alignment of the main Isar language elements. This depends on option
@{system_option_def "jedit_indent_newline"} (enabled by default).
Regular input (via keyboard or completion) indents the current line
whenever an new keyword is emerging at the start of the line. This depends
on option @{system_option_def "jedit_indent_input"} (enabled by default).
🚫[Semantic indentation] adds additional white space to unstructured proof
scripts via the number of subgoals. This requires information of ongoing
document processing and may thus lag behind when the user is editing too
quickly; see also option @{system_option_def "jedit_script_indent"} and
@{system_option_def "jedit_script_indent_limit"}.
The above options are accessible in the menu 🚫‹Plugins / Plugin Options /
Isabelle / General›. A prerequisite for advanced indentation is 🚫‹Utilities
/ Buffer Options / Automatic indentation›: it needs to be set to 🍋‹full›
(default).
›
section ‹SideKick parsers \label{sec:sidekick}›
text ‹
The 🚫‹SideKick› plugin provides some general services to display buffer
structure in a tree view. Isabelle/jEdit provides SideKick parsers for its
main mode for theory files, ML files, as well as some minor modes for the
🍋‹NEWS› file (see \figref{fig:sidekick}), session 🍋‹ROOT› files, system
🍋‹options›, and Bib{\TeX} files (\secref{sec:bibtex}).
\begin{figure}[!htb]
\begin{center}
\includegraphics[scale=0.333]{sidekick}
\end{center}
\caption{The Isabelle NEWS file with SideKick tree view}
\label{fig:sidekick}
\end{figure}
The default SideKick parser for theory files is 🍋‹isabelle›: it provides a
tree-view on the formal document structure, with section headings at the top
and formal specification elements at the bottom. The alternative parser
🍋‹isabelle-context› shows nesting of context blocks according to 🚫‹begin …
end› structure.
🚫
Isabelle/ML files are structured according to semi-formal comments that are
explained in 🍋‹"isabelle-implementation"›. This outline is turned into
a tree-view by default, by using the 🍋‹isabelle-ml› parser. There is also a
folding mode of the same name, for hierarchic text folds within ML files.
🚫
The special SideKick parser 🍋‹isabelle-markup› exposes the uninterpreted
markup tree of the PIDE document model of the current buffer. This is
occasionally useful for informative purposes, but the amount of displayed
information might cause problems for large buffers.
›
chapter ‹Prover IDE functionality \label{sec:document-model}›
section ‹Document model \label{sec:document-model}›
text ‹
The document model is central to the PIDE architecture: the editor and the
prover have a common notion of structured source text with markup, which is
produced by formal processing. The editor is responsible for edits of
document source, as produced by the user. The prover is responsible for
reports of document markup, as produced by its processing in the background.
Isabelle/jEdit handles classic editor events of jEdit, in order to connect
the physical world of the GUI (with its singleton state) to the mathematical
world of multiple document versions (with timeless and stateless updates).
›
subsection ‹Editor buffers and document nodes \label{sec:buffer-node}›
text ‹
As a regular text editor, jEdit maintains a collection of 🚫‹buffers› to
store text files; each buffer may be associated with any number of visible
🚫‹text areas›. Buffers are subject to an 🚫‹edit mode› that is determined
from the file name extension. The following modes are treated specifically
in Isabelle/jEdit:
🚫
\begin{tabular}{lll}
🚫‹mode› & 🚫‹file name› & 🚫‹content› \\\hline
🍋‹isabelle› & 🍋‹*.thy› & theory source \\
🍋‹isabelle-ml› & 🍋‹*.ML› & Isabelle/ML source \\
🍋‹sml› & 🍋‹*.sml› or 🍋‹*.sig› & Standard ML source \\
🍋‹isabelle-root› & 🍋‹ROOT› & session root \\
🍋‹isabelle-options› & & Isabelle options \\
🍋‹isabelle-news› & & Isabelle NEWS \\
\end{tabular}
🚫
All jEdit buffers are automatically added to the PIDE document-model as
🚫‹document nodes›. The overall document structure is defined by the theory
nodes in two dimensions:
🚫 via 🚫‹theory imports› that are specified in the 🚫‹theory header› using
concrete syntax of the @{command_ref theory} command 🍋‹"isabelle-isar-ref"›;
🚫 via 🚫‹auxiliary files› that are included into a theory by 🚫‹load
commands›, notably @{command_ref ML_file} and @{command_ref SML_file}
🍋‹"isabelle-isar-ref"›.
In any case, source files are managed by the PIDE infrastructure: the
physical file-system only plays a subordinate role. The relevant version of
source text is passed directly from the editor to the prover, using internal
communication channels.
›
subsection ‹Theories \label{sec:theories}›
text ‹
The 🚫‹Theories› panel (see also \figref{fig:theories}) provides an overview
of the status of continuous checking of theory nodes within the document
model.
\begin{figure}[!htb]
\begin{center}
\includegraphics[scale=0.333]{theories}
\end{center}
\caption{Theories panel with an overview of the document-model, and jEdit
text areas as editable views on some of the document nodes}
\label{fig:theories}
\end{figure}
Theory imports are resolved automatically by the PIDE document model: all
required files are loaded and stored internally, without the need to open
corresponding jEdit buffers. Opening or closing editor buffers later on has
no direct impact on the formal document content: it only affects visibility.
In contrast, auxiliary files (e.g.\ from @{command ML_file} commands) are
🚫‹not› resolved within the editor by default, but the prover process takes
care of that. This may be changed by enabling the system option
@{system_option jedit_auto_resolve}: it ensures that all files are uniformly
provided by the editor.
🚫
The visible 🚫‹perspective› of Isabelle/jEdit is defined by the collective
view on theory buffers via open text areas. The perspective is taken as a
hint for document processing: the prover ensures that those parts of a
theory where the user is looking are checked, while other parts that are
presently not required are ignored. The perspective is changed by opening or
closing text area windows, or scrolling within a window.
The 🚫‹Theories› panel provides some further options to influence the process
of continuous checking: it may be switched off globally to restrict the
prover to superficial processing of command syntax. It is also possible to
indicate theory nodes as 🚫‹required› for continuous checking: this means
such nodes and all their imports are always processed independently of the
visibility status (if continuous checking is enabled). Big theory libraries
that are marked as required can have significant impact on performance!
The 🚫‹Purge› button restricts the document model to theories that are
required for open editor buffers: inaccessible theories are removed and will
be rechecked when opened or imported later.
🚫
Formal markup of checked theory content is turned into GUI rendering, based
on a standard repertoire known from mainstream IDEs for programming
languages: colors, icons, highlighting, squiggly underlines, tooltips,
hyperlinks etc. For outer syntax of Isabelle/Isar there is some traditional
syntax-highlighting via static keywords and tokenization within the editor;
this buffer syntax is determined from theory imports. In contrast, the
painting of inner syntax (term language etc.)\ uses semantic information
that is reported dynamically from the logical context. Thus the prover can
provide additional markup to help the user to understand the meaning of
formal text, and to produce more text with some add-on tools (e.g.java.lang.NullPointerException
information messages with 🚫‹sendback› markup by automated provers or
disprovers in the background). ›
subsection ‹Auxiliary files \label{sec:aux-files}›
text ‹
Special load commands like @{command_ref ML_file} and @{command_ref
SML_file} 🍋‹"isabelle-isar-ref"› refer to auxiliary files within some
theory. Conceptually, the file argument of the command extends the theory
source by the content of the file, but its editor buffer may be loaded~/
changed~/ saved separately. The PIDE document model propagates changes of
auxiliary file content to the corresponding load command in the theory, to
update and process it accordingly: changes of auxiliary file content are
treated as changes of the corresponding load command.
🚫
As a concession to the massive amount of ML files in Isabelle/HOL itself,
the content of auxiliary files is only added to the PIDE document-model on
demand, the first time when opened explicitly in the editor. There are
further tricks to manage markup of ML files, such that Isabelle/HOL may be
edited conveniently in the Prover IDE on small machines with only 8\,GB of
main memory. Using 🍋‹Pure› as logic session image, the exploration may start
at the top 🍋‹$ISABELLE_HOME/src/HOL/Main.thy› or the bottom
🍋‹$ISABELLE_HOME/src/HOL/HOL.thy›, for example. It is also possible to
explore the Isabelle/Pure bootstrap process (a virtual copy) by opening
🍋‹$ISABELLE_HOME/src/Pure/ROOT.ML› like a theory in the Prover IDE.
Initially, before an auxiliary file is opened in the editor, the prover
reads its content from the physical file-system. After the file is opened
for the first time in the editor, e.g.\ by following the hyperlink
(\secref{sec:tooltips-hyperlinks}) for the argument of its @{command
ML_file} command, the content is taken from the jEdit buffer.
The change of responsibility from prover to editor counts as an update of
the document content, so subsequent theory sources need to be re-checked.
When the buffer is closed, the responsibility remains to the editor: the
file may be opened again without causing another document update.
A file that is opened in the editor, but its theory with the load command is
not, is presently inactive in the document model. A file that is loaded via
multiple load commands is associated to an arbitrary one: this situation is
morally unsupported and might lead to confusion.
🚫
Output that refers to an auxiliary file is combined with that of the
corresponding load command, and shown whenever the file or the command are
active (see also \secref{sec:output}).
Warnings, errors, and other useful markup is attached directly to the
positions in the auxiliary file buffer, in the manner of standard IDEs. By
using the load command @{command SML_file} as explained in
🍋‹$ISABELLE_HOME/src/Tools/SML/Examples.thy›, Isabelle/jEdit may be used as
fully-featured IDE for Standard ML, independently of theory or proof
development: the required theory merely serves as some kind of project file
for a collection of SML source modules.
›
section ‹Output \label{sec:output}›
text ‹
Prover output consists of 🚫‹markup› and 🚫‹messages›. Both are directly
attached to the corresponding positions in the original source text, and
visualized in the text area, e.g.\ as text colours for free and bound
variables, or as squiggly underlines for warnings, errors etc.\ (see also
\figref{fig:output}). In the latter case, the corresponding messages are
shown by hovering with the mouse over the highlighted text --- although in
many situations the user should already get some clue by looking at the
position of the text highlighting, without seeing the message body itself.
\begin{figure}[!htb]
\begin{center}
\includegraphics[scale=0.333]{output}
\end{center}
\caption{Multiple views on prover output: gutter with icon, text area with
popup, text overview column, 🚫‹Theories› panel, 🚫‹Output› panel}
\label{fig:output}
\end{figure}
The ``gutter'' on the left-hand-side of the text area uses icons to
provide a summary of the messages within the adjacent text line. Message
priorities are used to prefer errors over warnings, warnings over
information messages; other output is ignored.
The ``text overview column'' on the right-hand-side of the text area uses
similar information to paint small rectangles for the overall status of the
whole text buffer. The graphics is scaled to fit the logical buffer length
into the given window height. Mouse clicks on the overview area move the
cursor approximately to the corresponding text line in the buffer.
The 🚫‹Theories› panel provides another course-grained overview, but without
direct correspondence to text positions. The coloured rectangles represent
the amount of messages of a certain kind (warnings, errors, etc.) and the
execution status of commands. The border of each rectangle indicates the
overall status of processing: a thick border means it is 🚫‹finished› or
🚫‹failed› (with color for errors). A double-click on one of the theory
entries with their status overview opens the corresponding text buffer,
without moving the cursor to a specific point.
🚫
The 🚫‹Output› panel displays prover messages that correspond to a given
command, within a separate window. The cursor position in the presently
active text area determines the prover command whose cumulative message
output is appended and shown in that window (in canonical order according to
the internal execution of the command). There are also control elements to
modify the update policy of the output wrt.\ continued editor movements:
🚫‹Auto update› and 🚫‹Update›. This is particularly useful for multiple
instances of the 🚫‹Output› panel to look at different situations.
Alternatively, the panel can be turned into a passive 🚫‹Info› window via the
🚫‹Detach› menu item.
Proof state is handled separately (\secref{sec:state-output}), but it is
also possible to tick the corresponding checkbox to append it to regular
output (\figref{fig:output-including-state}). This is a globally persistent
option: it affects all open panels and future editor sessions.
\begin{figure}[!htb]
\begin{center}
\includegraphics[scale=0.333]{output-including-state}
\end{center}
\caption{Proof state display within the regular output panel}
\label{fig:output-including-state}
\end{figure}
🚫
🚫‹Explicit highlighting› of output works via the 🚫‹Search› field: it matches
the text against a given regular expression, in the notation of
Java.🚫‹🚫‹https://docs.oracle.com/en/java/javase/21/docs/api/java.base/java/util/regex/Pattern.html››
Results are also presented as a tree view, by sub-dividing the output panel
on demand.
🚫‹Implicit highlighting› of output is based on formal markup by the prover.
If the 🚫‹Auto hovering› option is enabled (default), then mouse hovering
alone is sufficient to see highlighted ranges stemming from nested syntax
structure (see also \secref{sec:tooltips-hyperlinks}); together with the ALT
keyboard modifier this produces a selection that is ready for copy-paste.
Without 🚫‹Auto hovering›, an additional keyboard modifier 🍋‹CONTROL› (Linux,
Windows) or 🍋‹COMMAND› (macOS) is required, as for input text.
🚫
Following the IDE principle, regular messages are attached to the original
source in the proper place and may be inspected on demand via popups. This
excludes messages that are somehow internal to the machinery of proof
checking, notably 🚫‹proof state› and 🚫‹tracing›.
In any case, the same display technology is used for small popups and big
output windows. The formal text contains markup that may be explored
recursively via further popups and hyperlinks (see
\secref{sec:tooltips-hyperlinks}), or clicked directly to initiate certain
actions (see \secref{sec:auto-tools} and \secref{sec:sledgehammer}).
🚫
Alternatively, the subsequent actions (with keyboard shortcuts) allow to
show tooltip messages or navigate error positions:
🚫
\begin{tabular}[t]{l}
@{action_ref "isabelle.tooltip"} (🍋‹CS+b›) \\
@{action_ref "isabelle.message"} (🍋‹CS+m›) \\
\end{tabular}\quad
\begin{tabular}[t]{l}
@{action_ref "isabelle.first-error"} (🍋‹CS+a›) \\
@{action_ref "isabelle.last-error"} (🍋‹CS+z›) \\
@{action_ref "isabelle.next-error"} (🍋‹CS+n›) \\
@{action_ref "isabelle.prev-error"} (🍋‹CS+p›) \\
\end{tabular}
🚫
›
section ‹Proof state \label{sec:state-output}›
text ‹
The main purpose of the Prover IDE is to help the user editing proof
documents, with ongoing formal checking by the prover in the background.
This can be done to some extent in the main text area alone, especially for
well-structured Isar proofs.
Nonetheless, internal proof state needs to be inspected in many situations
of exploration and ``debugging''. The 🚫‹State› panel shows exclusively such
proof state messages without further distraction, while all other messages
are displayed in 🚫‹Output› (\secref{sec:output}).
\Figref{fig:output-and-state} shows a typical GUI layout where both panels
are open, while the 🚫‹Proof state› option is disabled within 🚫‹Output›.
\begin{figure}[!htb]
\begin{center}
\includegraphics[scale=0.333]{output-and-state}
\end{center}
\caption{Separate proof state display (right) and other output (bottom).}
\label{fig:output-and-state}
\end{figure}
Another typical arrangement has more than one 🚫‹State› panel open (as
floating windows), with 🚫‹Auto update› disabled to look at an old situation
while the proof text in the vicinity is changed. The 🚫‹Update› button
triggers an explicit one-shot update; this operation is also available via
the action @{action "isabelle.update-state"} (keyboard shortcut 🍋‹S+ENTER›).
On small screens, it is occasionally useful to have all messages
concatenated in the regular 🚫‹Output› panel, e.g.\ see
\figref{fig:output-including-state}.
🚫
The mechanics of 🚫‹Output› versus 🚫‹State› are slightly different:
🚫 🚫‹Output› shows information that is continuously produced and already
present when the GUI wants to show it. This is implicitly controlled by
the visible perspective on the text.
🚫 🚫‹State› initiates a real-time query on demand, with a full round trip
including a fresh print operation on the prover side. This is controlled
explicitly when the cursor is moved to the next command (🚫‹Auto update›)
or the 🚫‹Update› operation is triggered.
This can make a difference in GUI responsibility and resource usage within
the prover process. Applications with very big proof states that are only
inspected in isolation work better with the 🚫‹State› panel.
›
section ‹Query \label{sec:query}›
text ‹
The 🚫‹Query› panel provides various GUI forms to request extra information
from the prover, as a replacement of old-style diagnostic commands like
--> --------------------
--> maximum size reached
--> --------------------