NEWS

more flexible session selection as in "isabelle jedit";

Isabelle NEWS -- history of user-relevant changes
=================================================

(Note: Isabelle/jEdit shows a tree-view of the NEWS file in Sidekick.)


New in Isabelle2018 (August 2018)
---------------------------------

*** General ***

* Session-qualified theory names are mandatory: it is no longer possible
to refer to unqualified theories from the parent session.
INCOMPATIBILITY for old developments that have not been updated to
Isabelle2017 yet (using the "isabelle imports" tool).

* Only the most fundamental theory names are global, usually the entry
points to major logic sessions: Pure, Main, Complex_Main, HOLCF, IFOL,
FOL, ZF, ZFC etc. INCOMPATIBILITY, need to use qualified names for
formerly global "HOL-Probability.Probability" and "HOL-SPARK.SPARK".

* Global facts need to be closed: no free variables and no hypotheses.
Rare INCOMPATIBILITY.

* Facts stemming from locale interpretation are subject to lazy
evaluation for improved performance. Rare INCOMPATIBILITY: errors
stemming from interpretation morphisms might be deferred and thus
difficult to locate; enable system option "strict_facts" temporarily to
avoid this.

* Marginal comments need to be written exclusively in the new-style form
"\<comment> \<open>text\<close>", old ASCII variants like "-- {* ... *}" are no longer
supported. INCOMPATIBILITY, use the command-line tool "isabelle
update_comments" to update existing theory files.

* Old-style inner comments (* ... *) within the term language are legacy
and will be discontinued soon: use formal comments "\<comment> \<open>...\<close>" or "\<^cancel>\<open>...\<close>"
instead.

* The "op <infix-op>" syntax for infix operators has been replaced by
"(<infix-op>)". If <infix-op> begins or ends with a "*", there needs to
be a space between the "*" and the corresponding parenthesis.
INCOMPATIBILITY, use the command-line tool "isabelle update_op" to
convert theory and ML files to the new syntax. Because it is based on
regular expression matching, the result may need a bit of manual
postprocessing. Invoking "isabelle update_op" converts all files in the
current directory (recursively). In case you want to exclude conversion
of ML files (because the tool frequently also converts ML's "op"
syntax), use option "-m".

* Theory header 'abbrevs' specifications need to be separated by 'and'.
INCOMPATIBILITY.

* Command 'external_file' declares the formal dependency on the given
file name, such that the Isabelle build process knows about it, but
without specific Prover IDE management.

* Session ROOT entries no longer allow specification of 'files'. Rare
INCOMPATIBILITY, use command 'external_file' within a proper theory
context.

* Session root directories may be specified multiple times: each
accessible ROOT file is processed only once. This facilitates
specification of $ISABELLE_HOME_USER/ROOTS or command-line options like
-d or -D for "isabelle build" and "isabelle jedit". Example:

  isabelle build -D '~~/src/ZF'

* The command 'display_drafts' has been discontinued. INCOMPATIBILITY,
use action "isabelle.draft" (or "print") in Isabelle/jEdit instead.

* In HTML output, the Isabelle symbol "\<hyphen>" is rendered as explicit
Unicode hyphen U+2010, to avoid unclear meaning of the old "soft hyphen"
U+00AD. Rare INCOMPATIBILITY, e.g. copy-paste of historic Isabelle HTML
output.


*** Isabelle/jEdit Prover IDE ***

* The command-line tool "isabelle jedit" provides more flexible options
for session management:

  - option -R builds an auxiliary logic image with all theories from
    other sessions that are not already present in its parent

  - option -S is like -R, with a focus on the selected session and its
    descendants (this reduces startup time for big projects like AFP)

  - option -A specifies an alternative ancestor session for options -R
    and -S

  - option -i includes additional sessions into the name-space of
    theories

  Examples:
    isabelle jedit -R HOL-Number_Theory
    isabelle jedit -R HOL-Number_Theory -A HOL
    isabelle jedit -d '$AFP' -S Formal_SSA -A HOL
    isabelle jedit -d '$AFP' -S Formal_SSA -A HOL-Analysis
    isabelle jedit -d '$AFP' -S Formal_SSA -A HOL-Analysis -i CryptHOL

* PIDE markup for session ROOT files: allows to complete session names,
follow links to theories and document files etc.

* Completion supports theory header imports, using theory base name.
E.g. "Prob" may be completed to "HOL-Probability.Probability".

* Named control symbols (without special Unicode rendering) are shown as
bold-italic keyword. This is particularly useful for the short form of
antiquotations with control symbol: \<^name>\<open>argument\<close>. The action
"isabelle.antiquoted_cartouche" turns an antiquotation with 0 or 1
arguments into this format.

* Completion provides templates for named symbols with arguments,
e.g. "\<comment> \<open>ARGUMENT\<close>" or "\<^emph>\<open>ARGUMENT\<close>".

* Slightly more parallel checking, notably for high priority print
functions (e.g. State output).

* The view title is set dynamically, according to the Isabelle
distribution and the logic session name. The user can override this via
set-view-title (stored persistently in $JEDIT_SETTINGS/perspective.xml).

* System options "spell_checker_include" and "spell_checker_exclude"
supersede former "spell_checker_elements" to determine regions of text
that are subject to spell-checking. Minor INCOMPATIBILITY.

* Action "isabelle.preview" is able to present more file formats,
notably bibtex database files and ML files.

* Action "isabelle.draft" is similar to "isabelle.preview", but shows a
plain-text document draft. Both are available via the menu "Plugins /
Isabelle".

* When loading text files, the Isabelle symbols encoding UTF-8-Isabelle
is only used if there is no conflict with existing Unicode sequences in
the file. Otherwise, the fallback encoding is plain UTF-8 and Isabelle
symbols remain in literal \<symbol> form. This avoids accidental loss of
Unicode content when saving the file.

* Bibtex database files (.bib) are semantically checked.

* Update to jedit-5.5.0, the latest release.


*** Isabelle/VSCode Prover IDE ***

* HTML preview of theories and other file-formats similar to
Isabelle/jEdit.

* Command-line tool "isabelle vscode_server" accepts the same options
-A, -R, -S, -i for session selection as "isabelle jedit". This is
relevant for isabelle.args configuration settings in VSCode. The former
option -A (explore all known session files) has been discontinued: it is
enabled by default, unless option -S is used to focus on a particular
spot in the session structure. INCOMPATIBILITY.


*** Document preparation ***

* Formal comments work uniformly in outer syntax, inner syntax (term
language), Isabelle/ML and some other embedded languages of Isabelle.
See also "Document comments" in the isar-ref manual. The following forms
are supported:

  - marginal text comment: \<comment> \<open>\<dots>\<close>
  - canceled source: \<^cancel>\<open>\<dots>\<close>
  - raw LaTeX: \<^latex>\<open>\<dots>\<close>

* Outside of the inner theory body, the default presentation context is
theory Pure. Thus elementary antiquotations may be used in markup
commands (e.g. 'chapter', 'section', 'text') and formal comments.

* System option "document_tags" specifies alternative command tags. This
is occasionally useful to control the global visibility of commands via
session options (e.g. in ROOT).

* Document markup commands ('section', 'text' etc.) are implicitly
tagged as "document" and visible by default. This avoids the application
of option "document_tags" to these commands.

* Isabelle names are mangled into LaTeX macro names to allow the full
identifier syntax with underscore, prime, digits. This is relevant for
antiquotations in control symbol notation, e.g. \<^const_name> becomes
\isactrlconstUNDERSCOREname.

* Document preparation with skip_proofs option now preserves the content
more accurately: only terminal proof steps ('by' etc.) are skipped.

* Document antiquotation @{theory name} requires the long
session-qualified theory name: this is what users reading the text
normally need to import.

* Document antiquotation @{session name} checks and prints the given
session name verbatim.

* Document antiquotation @{cite} now checks the given Bibtex entries
against the Bibtex database files -- only in batch-mode session builds.

* Command-line tool "isabelle document" has been re-implemented in
Isabelle/Scala, with simplified arguments and explicit errors from the
latex and bibtex process. Minor INCOMPATIBILITY.

* Session ROOT entry: empty 'document_files' means there is no document
for this session. There is no need to specify options [document = false]
anymore.


*** Isar ***

* Command 'interpret' no longer exposes resulting theorems as literal
facts, notably for the \<open>prop\<close> notation or the "fact" proof method. This
improves modularity of proofs and scalability of locale interpretation.
Rare INCOMPATIBILITY, need to refer to explicitly named facts instead
(e.g. use 'find_theorems' or 'try' to figure this out).

* The old 'def' command has been discontinued (legacy since
Isbelle2016-1). INCOMPATIBILITY, use 'define' instead -- usually with
object-logic equality or equivalence.


*** Pure ***

* The inner syntax category "sort" now includes notation "_" for the
dummy sort: it is effectively ignored in type-inference.

* Rewrites clauses (keyword 'rewrites') were moved into the locale
expression syntax, where they are part of locale instances. In
interpretation commands rewrites clauses now need to occur before 'for'
and 'defines'. Rare INCOMPATIBILITY; definitions immediately subject to
rewriting may need to be pulled up into the surrounding theory.

* For 'rewrites' clauses, if activating a locale instance fails, fall
back to reading the clause first. This helps avoid qualification of
locale instances where the qualifier's sole purpose is avoiding
duplicate constant declarations.

* Proof method "simp" now supports a new modifier "flip:" followed by a
list of theorems. Each of these theorems is removed from the simpset
(without warning if it is not there) and the symmetric version of the
theorem (i.e. lhs and rhs exchanged) is added to the simpset. For "auto"
and friends the modifier is "simp flip:".


*** HOL ***

* Sledgehammer: bundled version of "vampire" (for non-commercial users)
helps to avoid fragility of "remote_vampire" service.

* Clarified relationship of characters, strings and code generation:

  - Type "char" is now a proper datatype of 8-bit values.

  - Conversions "nat_of_char" and "char_of_nat" are gone; use more
    general conversions "of_char" and "char_of" with suitable type
    constraints instead.

  - The zero character is just written "CHR 0x00", not "0" any longer.

  - Type "String.literal" (for code generation) is now isomorphic to
    lists of 7-bit (ASCII) values; concrete values can be written as
    "STR ''...''" for sequences of printable characters and "STR 0x..."
    for one single ASCII code point given as hexadecimal numeral.

  - Type "String.literal" supports concatenation "... + ..." for all
    standard target languages.

  - Theory HOL-Library.Code_Char is gone; study the explanations
    concerning "String.literal" in the tutorial on code generation to
    get an idea how target-language string literals can be converted to
    HOL string values and vice versa.

  - Session Imperative-HOL: operation "raise" directly takes a value of
    type "String.literal" as argument, not type "string".

INCOMPATIBILITY.

* Code generation: Code generation takes an explicit option
"case_insensitive" to accomodate case-insensitive file systems.

* Abstract bit operations as part of Main: push_bit, take_bit, drop_bit.

* New, more general, axiomatization of complete_distrib_lattice. The
former axioms:

  "sup x (Inf X) = Inf (sup x ` X)" and "inf x (Sup X) = Sup (inf x ` X)"

are replaced by:

  "Inf (Sup ` A) <= Sup (Inf ` {f ` A | f . (! Y \<in> A . f Y \<in> Y)})"

The instantiations of sets and functions as complete_distrib_lattice are
moved to Hilbert_Choice.thy because their proofs need the Hilbert choice
operator. The dual of this property is also proved in theory
HOL.Hilbert_Choice.

* New syntax for the minimum/maximum of a function over a finite set:
MIN x\<in>A. B and even MIN x. B (only useful for finite types), also MAX.

* Clarifed theorem names:

  Min.antimono ~> Min.subset_imp
  Max.antimono ~> Max.subset_imp

Minor INCOMPATIBILITY.

* SMT module:

  - The 'smt_oracle' option is now necessary when using the 'smt' method
    with a solver other than Z3. INCOMPATIBILITY.

  - The encoding to first-order logic is now more complete in the
    presence of higher-order quantifiers. An 'smt_explicit_application'
    option has been added to control this. INCOMPATIBILITY.

* Facts sum.commute(_restrict) and prod.commute(_restrict) renamed to
sum.swap(_restrict) and prod.swap(_restrict), to avoid name clashes on
interpretation of abstract locales. INCOMPATIBILITY.

* Predicate coprime is now a real definition, not a mere abbreviation.
INCOMPATIBILITY.

* Predicate pairwise_coprime abolished, use "pairwise coprime" instead.
INCOMPATIBILITY.

* The relator rel_filter on filters has been strengthened to its
canonical categorical definition with better properties.
INCOMPATIBILITY.

* Generalized linear algebra involving linear, span, dependent, dim
from type class real_vector to locales module and vector_space.
Renamed:

  span_inc ~> span_superset
  span_superset ~> span_base
  span_eq ~> span_eq_iff

INCOMPATIBILITY.

* Class linordered_semiring_1 covers zero_less_one also, ruling out
pathologic instances. Minor INCOMPATIBILITY.

* Theory HOL.List: functions "sorted_wrt" and "sorted" now compare every
element in a list to all following elements, not just the next one.

* Theory HOL.List syntax:

  - filter-syntax "[x <- xs. P]" is no longer output syntax, but only
    input syntax

  - list comprehension syntax now supports tuple patterns in "pat <- xs"

* Theory Map: "empty" must now be qualified as "Map.empty".

* Removed nat-int transfer machinery. Rare INCOMPATIBILITY.

* Fact mod_mult_self4 (on nat) renamed to Suc_mod_mult_self3, to avoid
clash with fact mod_mult_self4 (on more generic semirings).
INCOMPATIBILITY.

* Eliminated some theorem aliasses:
  even_times_iff ~> even_mult_iff
  mod_2_not_eq_zero_eq_one_nat ~> not_mod_2_eq_0_eq_1
  even_of_nat ~> even_int_iff

INCOMPATIBILITY.

* Eliminated some theorem duplicate variations:

  - dvd_eq_mod_eq_0_numeral can be replaced by dvd_eq_mod_eq_0
  - mod_Suc_eq_Suc_mod can be replaced by mod_Suc
  - mod_Suc_eq_Suc_mod [symmetrict] can be replaced by mod_simps
  - mod_eq_0_iff can be replaced by mod_eq_0_iff_dvd and dvd_def
  - the witness of mod_eqD can be given directly as "_ div _"

INCOMPATIBILITY.

* Classical setup: Assumption "m mod d = 0" (for m d :: nat) is no
longer aggresively destroyed to "\<exists>q. m = d * q". INCOMPATIBILITY, adding
"elim!: dvd" to classical proof methods in most situations restores
broken proofs.

* Theory HOL-Library.Conditional_Parametricity provides command
'parametric_constant' for proving parametricity of non-recursive
definitions. For constants that are not fully parametric the command
will infer conditions on relations (e.g., bi_unique, bi_total, or type
class conditions such as "respects 0") sufficient for parametricity. See
theory HOL-ex.Conditional_Parametricity_Examples for some examples.

* Theory HOL-Library.Code_Lazy provides a new preprocessor for the code
generator to generate code for algebraic types with lazy evaluation
semantics even in call-by-value target languages. See the theories
HOL-ex.Code_Lazy_Demo and HOL-Codegenerator_Test.Code_Lazy_Test for some
examples.

* Theory HOL-Library.Landau_Symbols has been moved here from AFP.

* Theory HOL-Library.Old_Datatype no longer provides the legacy command
'old_datatype'. INCOMPATIBILITY.

* Theory HOL-Computational_Algebra.Polynomial_Factorial does not provide
instances of rat, real, complex as factorial rings etc. Import
HOL-Computational_Algebra.Field_as_Ring explicitly in case of need.
INCOMPATIBILITY.

* Session HOL-Algebra: renamed (^) to [^] to avoid conflict with new
infix/prefix notation.

* Session HOL-Algebra: revamped with much new material. The set of
isomorphisms between two groups is now denoted iso rather than iso_set.
INCOMPATIBILITY.

* Session HOL-Analysis: the Arg function now respects the same interval
as Ln, namely (-pi,pi]; the old Arg function has been renamed Arg2pi.
INCOMPATIBILITY.

* Session HOL-Analysis: the functions zorder, zer_poly, porder and
pol_poly have been redefined. All related lemmas have been reworked.
INCOMPATIBILITY.

* Session HOL-Analysis: infinite products, Moebius functions, the
Riemann mapping theorem, the Vitali covering theorem,
change-of-variables results for integration and measures.

* Session HOL-Real_Asymp: proof method "real_asymp" proves asymptotics
or real-valued functions (limits, "Big-O", etc.) automatically.
See also ~~/src/HOL/Real_Asymp/Manual for some documentation.

* Session HOL-Types_To_Sets: more tool support (unoverload_type combines
internalize_sorts and unoverload) and larger experimental application
(type based linear algebra transferred to linear algebra on subspaces).


*** ML ***

* Operation Export.export emits theory exports (arbitrary blobs), which
are stored persistently in the session build database.

* Command 'ML_export' exports ML toplevel bindings to the global
bootstrap environment of the ML process. This allows ML evaluation
without a formal theory context, e.g. in command-line tools like
"isabelle process".


*** System ***

* Mac OS X 10.10 Yosemite is now the baseline version; Mavericks is no
longer supported.

* Linux and Windows/Cygwin is for x86_64 only, old 32bit platform
support has been discontinued.

* Java runtime is for x86_64 only. Corresponding Isabelle settings have
been renamed to ISABELLE_TOOL_JAVA_OPTIONS and JEDIT_JAVA_OPTIONS,
instead of former 32/64 variants. INCOMPATIBILITY.

* Old settings ISABELLE_PLATFORM and ISABELLE_WINDOWS_PLATFORM should be
phased out due to unclear preference of 32bit vs. 64bit architecture.
Explicit GNU bash expressions are now preferred, for example (with
quotes):

  #Posix executables (Unix or Cygwin), with preference for 64bit
  "${ISABELLE_PLATFORM64:-$ISABELLE_PLATFORM32}"

  #native Windows or Unix executables, with preference for 64bit
  "${ISABELLE_WINDOWS_PLATFORM64:-${ISABELLE_WINDOWS_PLATFORM32:-${ISABELLE_PLATFORM64:-$ISABELLE_PLATFORM32}}}"

  #native Windows (32bit) or Unix executables (preference for 64bit)
  "${ISABELLE_WINDOWS_PLATFORM32:-${ISABELLE_PLATFORM64:-$ISABELLE_PLATFORM32}}"

* Command-line tool "isabelle build" supports new options:
  - option -B NAME: include session NAME and all descendants
  - option -S: only observe changes of sources, not heap images
  - option -f: forces a fresh build

* Command-line tool "isabelle build" takes "condition" options with the
corresponding environment values into account, when determining the
up-to-date status of a session.

* The command-line tool "dump" dumps information from the cumulative
PIDE session database: many sessions may be loaded into a given logic
image, results from all loaded theories are written to the output
directory.

* Command-line tool "isabelle imports -I" also reports actual session
imports. This helps to minimize the session dependency graph.

* The command-line tool "export" and 'export_files' in session ROOT
entries retrieve theory exports from the session build database.

* The command-line tools "isabelle server" and "isabelle client" provide
access to the Isabelle Server: it supports responsive session management
and concurrent use of theories, based on Isabelle/PIDE infrastructure.
See also the "system" manual.

* The command-line tool "isabelle update_comments" normalizes formal
comments in outer syntax as follows: \<comment> \<open>text\<close> (whith a single space to
approximate the appearance in document output). This is more specific
than former "isabelle update_cartouches -c": the latter tool option has
been discontinued.

* The command-line tool "isabelle mkroot" now always produces a document
outline: its options have been adapted accordingly. INCOMPATIBILITY.

* The command-line tool "isabelle mkroot -I" initializes a Mercurial
repository for the generated session files.

* Settings ISABELLE_HEAPS + ISABELLE_BROWSER_INFO (or
ISABELLE_HEAPS_SYSTEM + ISABELLE_BROWSER_INFO_SYSTEM in "system build
mode") determine the directory locations of the main build artefacts --
instead of hard-wired directories in ISABELLE_HOME_USER (or
ISABELLE_HOME).

* Settings ISABELLE_PATH and ISABELLE_OUTPUT have been discontinued:
heap images and session databases are always stored in
$ISABELLE_HEAPS/$ML_IDENTIFIER (command-line default) or
$ISABELLE_HEAPS_SYSTEM/$ML_IDENTIFIER (main Isabelle application or
"isabelle jedit -s" or "isabelle build -s").

* ISABELLE_LATEX and ISABELLE_PDFLATEX now include platform-specific
options for improved error reporting. Potential INCOMPATIBILITY with
unusual LaTeX installations, may have to adapt these settings.

* Update to Poly/ML 5.7.1 with slightly improved performance and PIDE
markup for identifier bindings. It now uses The GNU Multiple Precision
Arithmetic Library (libgmp) on all platforms, notably Mac OS X with
32/64 bit.



New in Isabelle2017 (October 2017)
----------------------------------

*** General ***

* Experimental support for Visual Studio Code (VSCode) as alternative
Isabelle/PIDE front-end, see also
https://marketplace.visualstudio.com/items?itemName=makarius.Isabelle2017

VSCode is a new type of application that continues the concepts of
"programmer's editor" and "integrated development environment" towards
fully semantic editing and debugging -- in a relatively light-weight
manner. Thus it fits nicely on top of the Isabelle/PIDE infrastructure.
Technically, VSCode is based on the Electron application framework
(Node.js + Chromium browser + V8), which is implemented in JavaScript
and TypeScript, while Isabelle/VSCode mainly consists of Isabelle/Scala
modules around a Language Server implementation.

* Theory names are qualified by the session name that they belong to.
This affects imports, but not the theory name space prefix (which is
just the theory base name as before).

In order to import theories from other sessions, the ROOT file format
provides a new 'sessions' keyword. In contrast, a theory that is
imported in the old-fashioned manner via an explicit file-system path
belongs to the current session, and might cause theory name conflicts
later on. Theories that are imported from other sessions are excluded
from the current session document. The command-line tool "isabelle
imports" helps to update theory imports.

* The main theory entry points for some non-HOL sessions have changed,
to avoid confusion with the global name "Main" of the session HOL. This
leads to the follow renamings:

  CTT/Main.thy    ~>  CTT/CTT.thy
  ZF/Main.thy     ~>  ZF/ZF.thy
  ZF/Main_ZF.thy  ~>  ZF/ZF.thy
  ZF/Main_ZFC.thy ~>  ZF/ZFC.thy
  ZF/ZF.thy       ~>  ZF/ZF_Base.thy

INCOMPATIBILITY.

* Commands 'alias' and 'type_alias' introduce aliases for constants and
type constructors, respectively. This allows adhoc changes to name-space
accesses within global or local theory contexts, e.g. within a 'bundle'.

* Document antiquotations @{prf} and @{full_prf} output proof terms
(again) in the same way as commands 'prf' and 'full_prf'.

* Computations generated by the code generator can be embedded directly
into ML, alongside with @{code} antiquotations, using the following
antiquotations:

  @{computation ... terms: ... datatypes: ...} :
    ((term -> term) -> 'ml option -> 'a) -> Proof.context -> term -> 'a
  @{computation_conv ... terms: ... datatypes: ...} :
    (Proof.context -> 'ml -> conv) -> Proof.context -> conv
  @{computation_check terms: ... datatypes: ...} : Proof.context -> conv

See src/HOL/ex/Computations.thy,
src/HOL/Decision_Procs/Commutative_Ring.thy and
src/HOL/Decision_Procs/Reflective_Field.thy for examples and the
tutorial on code generation.


*** Prover IDE -- Isabelle/Scala/jEdit ***

* Session-qualified theory imports allow the Prover IDE to process
arbitrary theory hierarchies independently of the underlying logic
session image (e.g. option "isabelle jedit -l"), but the directory
structure needs to be known in advance (e.g. option "isabelle jedit -d"
or a line in the file $ISABELLE_HOME_USER/ROOTS).

* The PIDE document model maintains file content independently of the
status of jEdit editor buffers. Reloading jEdit buffers no longer causes
changes of formal document content. Theory dependencies are always
resolved internally, without the need for corresponding editor buffers.
The system option "jedit_auto_load" has been discontinued: it is
effectively always enabled.

* The Theories dockable provides a "Purge" button, in order to restrict
the document model to theories that are required for open editor
buffers.

* The Theories dockable indicates the overall status of checking of each
entry. When all forked tasks of a theory are finished, the border is
painted with thick lines; remaining errors in this situation are
represented by a different border color.

* Automatic indentation is more careful to avoid redundant spaces in
intermediate situations. Keywords are indented after input (via typed
characters or completion); see also option "jedit_indent_input".

* Action "isabelle.preview" opens an HTML preview of the current theory
document in the default web browser.

* Command-line invocation "isabelle jedit -R -l LOGIC" opens the ROOT
entry of the specified logic session in the editor, while its parent is
used for formal checking.

* The main Isabelle/jEdit plugin may be restarted manually (using the
jEdit Plugin Manager), as long as the "Isabelle Base" plugin remains
enabled at all times.

* Update to current jedit-5.4.0.


*** Pure ***

* Deleting the last code equations for a particular function using
[code del] results in function with no equations (runtime abort) rather
than an unimplemented function (generation time abort). Use explicit
[[code drop:]] to enforce the latter. Minor INCOMPATIBILITY.

* Proper concept of code declarations in code.ML:
  - Regular code declarations act only on the global theory level, being
    ignored with warnings if syntactically malformed.
  - Explicitly global code declarations yield errors if syntactically
    malformed.
  - Default code declarations are silently ignored if syntactically
    malformed.
Minor INCOMPATIBILITY.

* Clarified and standardized internal data bookkeeping of code
declarations: history of serials allows to track potentially
non-monotonous declarations appropriately. Minor INCOMPATIBILITY.


*** HOL ***

* The Nunchaku model finder is now part of "Main".

* SMT module:
  - A new option, 'smt_nat_as_int', has been added to translate 'nat' to
    'int' and benefit from the SMT solver's theory reasoning. It is
    disabled by default.
  - The legacy module "src/HOL/Library/Old_SMT.thy" has been removed.
  - Several small issues have been rectified in the 'smt' command.

* (Co)datatype package: The 'size_gen_o_map' lemma is no longer
generated for datatypes with type class annotations. As a result, the
tactic that derives it no longer fails on nested datatypes. Slight
INCOMPATIBILITY.

* Command and antiquotation "value" with modified default strategy:
terms without free variables are always evaluated using plain evaluation
only, with no fallback on normalization by evaluation. Minor
INCOMPATIBILITY.

* Theories "GCD" and "Binomial" are already included in "Main" (instead
of "Complex_Main").

* Constant "surj" is a full input/output abbreviation (again).
Minor INCOMPATIBILITY.

* Dropped aliasses RangeP, DomainP for Rangep, Domainp respectively.
INCOMPATIBILITY.

* Renamed ii to imaginary_unit in order to free up ii as a variable
name. The syntax \<i> remains available. INCOMPATIBILITY.

* Dropped abbreviations transP, antisymP, single_valuedP; use constants
transp, antisymp, single_valuedp instead. INCOMPATIBILITY.

* Constant "subseq" in Topological_Spaces has been removed -- it is
subsumed by "strict_mono". Some basic lemmas specific to "subseq" have
been renamed accordingly, e.g. "subseq_o" -> "strict_mono_o" etc.

* Theory List: "sublist" renamed to "nths" in analogy with "nth", and
"sublisteq" renamed to "subseq". Minor INCOMPATIBILITY.

* Theory List: new generic function "sorted_wrt".

* Named theorems mod_simps covers various congruence rules concerning
mod, replacing former zmod_simps. INCOMPATIBILITY.

* Swapped orientation of congruence rules mod_add_left_eq,
mod_add_right_eq, mod_add_eq, mod_mult_left_eq, mod_mult_right_eq,
mod_mult_eq, mod_minus_eq, mod_diff_left_eq, mod_diff_right_eq,
mod_diff_eq. INCOMPATIBILITY.

* Generalized some facts:
    measure_induct_rule
    measure_induct
    zminus_zmod ~> mod_minus_eq
    zdiff_zmod_left ~> mod_diff_left_eq
    zdiff_zmod_right ~> mod_diff_right_eq
    zmod_eq_dvd_iff ~> mod_eq_dvd_iff
INCOMPATIBILITY.

* Algebraic type class hierarchy of euclidean (semi)rings in HOL:
euclidean_(semi)ring, euclidean_(semi)ring_cancel,
unique_euclidean_(semi)ring; instantiation requires provision of a
euclidean size.

* Theory "HOL-Number_Theory.Euclidean_Algorithm" has been reworked:
  - Euclidean induction is available as rule eucl_induct.
  - Constants Euclidean_Algorithm.gcd, Euclidean_Algorithm.lcm,
    Euclidean_Algorithm.Gcd and Euclidean_Algorithm.Lcm allow
    easy instantiation of euclidean (semi)rings as GCD (semi)rings.
  - Coefficients obtained by extended euclidean algorithm are
    available as "bezout_coefficients".
INCOMPATIBILITY.

* Theory "Number_Theory.Totient" introduces basic notions about Euler's
totient function previously hidden as solitary example in theory
Residues. Definition changed so that "totient 1 = 1" in agreement with
the literature. Minor INCOMPATIBILITY.

* New styles in theory "HOL-Library.LaTeXsugar":
  - "dummy_pats" for printing equations with "_" on the lhs;
  - "eta_expand" for printing eta-expanded terms.

* Theory "HOL-Library.Permutations": theorem bij_swap_ompose_bij has
been renamed to bij_swap_compose_bij. INCOMPATIBILITY.

* New theory "HOL-Library.Going_To_Filter" providing the "f going_to F"
filter for describing points x such that f(x) is in the filter F.

* Theory "HOL-Library.Formal_Power_Series": constants X/E/L/F have been
renamed to fps_X/fps_exp/fps_ln/fps_hypergeo to avoid polluting the name
space. INCOMPATIBILITY.

* Theory "HOL-Library.FinFun" has been moved to AFP (again).
INCOMPATIBILITY.

* Theory "HOL-Library.FuncSet": some old and rarely used ASCII
replacement syntax has been removed. INCOMPATIBILITY, standard syntax
with symbols should be used instead. The subsequent commands help to
reproduce the old forms, e.g. to simplify porting old theories:

syntax (ASCII)
  "_PiE" :: "pttrn \<Rightarrow> 'a set \<Rightarrow> 'b set \<Rightarrow> ('a \<Rightarrow> 'b) set"  ("(3PIE _:_./ _)" 10)
  "_Pi"  :: "pttrn \<Rightarrow> 'a set \<Rightarrow> 'b set \<Rightarrow> ('a \<Rightarrow> 'b) set"  ("(3PI _:_./ _)" 10)
  "_lam" :: "pttrn \<Rightarrow> 'a set \<Rightarrow> 'a \<Rightarrow> 'b \<Rightarrow> ('a \<Rightarrow> 'b)"  ("(3%_:_./ _)" [0,0,3] 3)

* Theory "HOL-Library.Multiset": the simprocs on subsets operators of
multisets have been renamed:

  msetless_cancel_numerals ~> msetsubset_cancel
  msetle_cancel_numerals ~> msetsubset_eq_cancel

INCOMPATIBILITY.

* Theory "HOL-Library.Pattern_Aliases" provides input and output syntax
for pattern aliases as known from Haskell, Scala and ML.

* Theory "HOL-Library.Uprod" formalizes the type of unordered pairs.

* Session HOL-Analysis: more material involving arcs, paths, covering
spaces, innessential maps, retracts, infinite products, simplicial
complexes. Baire Category theorem. Major results include the Jordan
Curve Theorem and the Great Picard Theorem.

* Session HOL-Algebra has been extended by additional lattice theory:
the Knaster-Tarski fixed point theorem and Galois Connections.

* Sessions HOL-Computational_Algebra and HOL-Number_Theory: new notions
of squarefreeness, n-th powers, and prime powers.

* Session "HOL-Computional_Algebra" covers many previously scattered
theories, notably Euclidean_Algorithm, Factorial_Ring,
Formal_Power_Series, Fraction_Field, Fundamental_Theorem_Algebra,
Normalized_Fraction, Polynomial_FPS, Polynomial, Primes. Minor
INCOMPATIBILITY.


*** System ***

* Isabelle/Scala: the SQL module supports access to relational
databases, either as plain file (SQLite) or full-scale server
(PostgreSQL via local port or remote ssh connection).

* Results of "isabelle build" are recorded as SQLite database (i.e.
"Application File Format" in the sense of
https://www.sqlite.org/appfileformat.html). This allows systematic
access via operations from module Sessions.Store in Isabelle/Scala.

* System option "parallel_proofs" is 1 by default (instead of more
aggressive 2). This requires less heap space and avoids burning parallel
CPU cycles, while full subproof parallelization is enabled for repeated
builds (according to parallel_subproofs_threshold).

* System option "record_proofs" allows to change the global
Proofterm.proofs variable for a session. Regular values are are 0, 1, 2;
a negative value means the current state in the ML heap image remains
unchanged.

* Isabelle settings variable ISABELLE_SCALA_BUILD_OPTIONS has been
renamed to ISABELLE_SCALAC_OPTIONS. Rare INCOMPATIBILITY.

* Isabelle settings variables ISABELLE_WINDOWS_PLATFORM,
ISABELLE_WINDOWS_PLATFORM32, ISABELLE_WINDOWS_PLATFORM64 indicate the
native Windows platform (independently of the Cygwin installation). This
is analogous to ISABELLE_PLATFORM, ISABELLE_PLATFORM32,
ISABELLE_PLATFORM64.

* Command-line tool "isabelle build_docker" builds a Docker image from
the Isabelle application bundle for Linux. See also
https://hub.docker.com/r/makarius/isabelle

* Command-line tool "isabelle vscode_server" provides a Language Server
Protocol implementation, e.g. for the Visual Studio Code editor. It
serves as example for alternative PIDE front-ends.

* Command-line tool "isabelle imports" helps to maintain theory imports
wrt. session structure. Examples for the main Isabelle distribution:

  isabelle imports -I -a
  isabelle imports -U -a
  isabelle imports -U -i -a
  isabelle imports -M -a -d '~~/src/Benchmarks'



New in Isabelle2016-1 (December 2016)
-------------------------------------

*** General ***

* Splitter in proof methods "simp", "auto" and friends:
  - The syntax "split add" has been discontinued, use plain "split",
    INCOMPATIBILITY.
  - For situations with many conditional or case expressions, there is
    an alternative splitting strategy that can be much faster. It is
    selected by writing "split!" instead of "split". It applies safe
    introduction and elimination rules after each split rule. As a
    result the subgoal may be split into several subgoals.

* Command 'bundle' provides a local theory target to define a bundle
from the body of specification commands (such as 'declare',
'declaration', 'notation', 'lemmas', 'lemma'). For example:

bundle foo
begin
  declare a [simp]
  declare b [intro]
end

* Command 'unbundle' is like 'include', but works within a local theory
context. Unlike "context includes ... begin", the effect of 'unbundle'
on the target context persists, until different declarations are given.

* Simplified outer syntax: uniform category "name" includes long
identifiers. Former "xname" / "nameref" / "name reference" has been
discontinued.

* Embedded content (e.g. the inner syntax of types, terms, props) may be
delimited uniformly via cartouches. This works better than old-fashioned
quotes when sub-languages are nested.

* Mixfix annotations support general block properties, with syntax
"(\<open>x=a y=b z \<dots>\<close>". Notable property names are "indent", "consistent",
"unbreakable", "markup". The existing notation "(DIGITS" is equivalent
to "(\<open>indent=DIGITS\<close>". The former notation "(00" for unbreakable blocks
is superseded by "(\<open>unbreabable\<close>" --- rare INCOMPATIBILITY.

* Proof method "blast" is more robust wrt. corner cases of Pure
statements without object-logic judgment.

* Commands 'prf' and 'full_prf' are somewhat more informative (again):
proof terms are reconstructed and cleaned from administrative thm nodes.

* Code generator: config option "code_timing" triggers measurements of
different phases of code generation. See src/HOL/ex/Code_Timing.thy for
examples.

* Code generator: implicits in Scala (stemming from type class
instances) are generated into companion object of corresponding type
class, to resolve some situations where ambiguities may occur.

* Solve direct: option "solve_direct_strict_warnings" gives explicit
warnings for lemma statements with trivial proofs.


*** Prover IDE -- Isabelle/Scala/jEdit ***

* More aggressive flushing of machine-generated input, according to
system option editor_generated_input_delay (in addition to existing
editor_input_delay for regular user edits). This may affect overall PIDE
reactivity and CPU usage.

* Syntactic indentation according to Isabelle outer syntax. Action
"indent-lines" (shortcut C+i) indents the current line according to
command keywords and some command substructure. Action
"isabelle.newline" (shortcut ENTER) indents the old and the new line
according to command keywords only; see also option
"jedit_indent_newline".

* Semantic indentation for unstructured proof scripts ('apply' etc.) via
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 "jedit_script_indent" and
"jedit_script_indent_limit".

* Refined folding mode "isabelle" based on Isar syntax: 'next' and 'qed'
are treated as delimiters for fold structure; 'begin' and 'end'
structure of theory specifications is treated as well.

* Command 'proof' provides information about proof outline with cases,
e.g. for proof methods "cases", "induct", "goal_cases".

* Completion templates for commands involving "begin ... end" blocks,
e.g. 'context', 'notepad'.

* Sidekick parser "isabelle-context" shows nesting of context blocks
according to 'begin' and 'end' structure.

* Highlighting of entity def/ref positions wrt. cursor.

* Action "isabelle.select-entity" (shortcut CS+ENTER) selects all
occurrences of the formal entity at the caret position. This facilitates
systematic renaming.

* PIDE document markup works across multiple Isar commands, e.g. the
results established at the end of a proof are properly identified in the
theorem statement.

* Cartouche abbreviations work both for " and ` to accomodate typical
situations where old ASCII notation may be updated.

* Dockable window "Symbols" also provides access to 'abbrevs' from the
outer syntax of the current theory buffer. This provides clickable
syntax templates, including entries with empty abbrevs name (which are
inaccessible via keyboard completion).

* IDE support for the Isabelle/Pure bootstrap process, with the
following independent stages:

  src/Pure/ROOT0.ML
  src/Pure/ROOT.ML
  src/Pure/Pure.thy
  src/Pure/ML_Bootstrap.thy

The ML ROOT files act like quasi-theories in the context of theory
ML_Bootstrap: this allows continuous checking of all loaded ML files.
The theory files are presented with a modified header to import Pure
from the running Isabelle instance. Results from changed versions of
each stage are *not* propagated to the next stage, and isolated from the
actual Isabelle/Pure that runs the IDE itself. The sequential
dependencies of the above files are only observed for batch build.

* Isabelle/ML and Standard ML files are presented in Sidekick with the
tree structure of section headings: this special comment format is
described in "implementation" chapter 0, e.g. (*** section ***).

* Additional abbreviations for syntactic completion may be specified
within the theory header as 'abbrevs'. The theory syntax for 'keywords'
has been simplified accordingly: optional abbrevs need to go into the
new 'abbrevs' section.

* Global abbreviations via $ISABELLE_HOME/etc/abbrevs and
$ISABELLE_HOME_USER/etc/abbrevs are no longer supported. Minor
INCOMPATIBILITY, use 'abbrevs' within theory header instead.

* Action "isabelle.keymap-merge" asks the user to resolve pending
Isabelle keymap changes that are in conflict with the current jEdit
keymap; non-conflicting changes are always applied implicitly. This
action is automatically invoked on Isabelle/jEdit startup and thus
increases chances that users see new keyboard shortcuts when re-using
old keymaps.

* ML and document antiquotations for file-systems paths are more uniform
and diverse:

  @{path NAME}   -- no file-system check
  @{file NAME}   -- check for plain file
  @{dir NAME}    -- check for directory

Minor INCOMPATIBILITY, former uses of @{file} and @{file_unchecked} may
have to be changed.


*** Document preparation ***

* New symbol \<circle>, e.g. for temporal operator.

* New document and ML antiquotation @{locale} for locales, similar to
existing antiquotation @{class}.

* Mixfix annotations support delimiters like \<^control>\<open>cartouche\<close> --
this allows special forms of document output.

* Raw LaTeX output now works via \<^latex>\<open>...\<close> instead of raw control
symbol \<^raw:...>. INCOMPATIBILITY, notably for LaTeXsugar.thy and its
derivatives.

* \<^raw:...> symbols are no longer supported.

* Old 'header' command is no longer supported (legacy since
Isabelle2015).


*** Isar ***

* Many specification elements support structured statements with 'if' /
'for' eigen-context, e.g. 'axiomatization', 'abbreviation',
'definition', 'inductive', 'function'.

* Toplevel theorem statements support eigen-context notation with 'if' /
'for' (in postfix), which corresponds to 'assumes' / 'fixes' in the
traditional long statement form (in prefix). Local premises are called
"that" or "assms", respectively. Empty premises are *not* bound in the
context: INCOMPATIBILITY.

* Command 'define' introduces a local (non-polymorphic) definition, with
optional abstraction over local parameters. The syntax resembles
'definition' and 'obtain'. It fits better into the Isar language than
old 'def', which is now a legacy feature.

* Command 'obtain' supports structured statements with 'if' / 'for'
context.

* Command '\<proof>' is an alias for 'sorry', with different
typesetting. E.g. to produce proof holes in examples and documentation.

* The defining position of a literal fact \<open>prop\<close> is maintained more
carefully, and made accessible as hyperlink in the Prover IDE.

* Commands 'finally' and 'ultimately' used to expose the result as
literal fact: this accidental behaviour has been discontinued. Rare
INCOMPATIBILITY, use more explicit means to refer to facts in Isar.

* Command 'axiomatization' has become more restrictive to correspond
better to internal axioms as singleton facts with mandatory name. Minor
INCOMPATIBILITY.

* Proof methods may refer to the main facts via the dynamic fact
"method_facts". This is particularly useful for Eisbach method
definitions.

* Proof method "use" allows to modify the main facts of a given method
expression, e.g.

  (use facts in simp)
  (use facts in \<open>simp add: ...\<close>)

* The old proof method "default" has been removed (legacy since
Isabelle2016). INCOMPATIBILITY, use "standard" instead.


*** Pure ***

* Pure provides basic versions of proof methods "simp" and "simp_all"
that only know about meta-equality (==). Potential INCOMPATIBILITY in
theory imports that merge Pure with e.g. Main of Isabelle/HOL: the order
is relevant to avoid confusion of Pure.simp vs. HOL.simp.

* The command 'unfolding' and proof method "unfold" include a second
stage where given equations are passed through the attribute "abs_def"
before rewriting. This ensures that definitions are fully expanded,
regardless of the actual parameters that are provided. Rare
INCOMPATIBILITY in some corner cases: use proof method (simp only:)
instead, or declare [[unfold_abs_def = false]] in the proof context.

* Type-inference improves sorts of newly introduced type variables for
the object-logic, using its base sort (i.e. HOL.type for Isabelle/HOL).
Thus terms like "f x" or "\<And>x. P x" without any further syntactic context
produce x::'a::type in HOL instead of x::'a::{} in Pure. Rare
INCOMPATIBILITY, need to provide explicit type constraints for Pure
types where this is really intended.


*** HOL ***

* New proof method "argo" using the built-in Argo solver based on SMT
technology. The method can be used to prove goals of quantifier-free
propositional logic, goals based on a combination of quantifier-free
propositional logic with equality, and goals based on a combination of
quantifier-free propositional logic with linear real arithmetic
including min/max/abs. See HOL/ex/Argo_Examples.thy for examples.

* The new "nunchaku" command integrates the Nunchaku model finder. The
tool is experimental. See ~~/src/HOL/Nunchaku/Nunchaku.thy for details.

* Metis: The problem encoding has changed very slightly. This might
break existing proofs. INCOMPATIBILITY.

* Sledgehammer:
  - The MaSh relevance filter is now faster than before.
  - Produce syntactically correct Vampire 4.0 problem files.

* (Co)datatype package:
  - New commands for defining corecursive functions and reasoning about
    them in "~~/src/HOL/Library/BNF_Corec.thy": 'corec', 'corecursive',
    'friend_of_corec', and 'corecursion_upto'; and 'corec_unique' proof
    method. See 'isabelle doc corec'.
  - The predicator :: ('a \<Rightarrow> bool) \<Rightarrow> 'a F \<Rightarrow> bool is now a first-class
    citizen in bounded natural functors.
  - 'primrec' now allows nested calls through the predicator in addition
    to the map function.
  - 'bnf' automatically discharges reflexive proof obligations.
  - 'bnf' outputs a slightly modified proof obligation expressing rel in
       terms of map and set
       (not giving a specification for rel makes this one reflexive).
  - 'bnf' outputs a new proof obligation expressing pred in terms of set
       (not giving a specification for pred makes this one reflexive).
    INCOMPATIBILITY: manual 'bnf' declarations may need adjustment.
  - Renamed lemmas:
      rel_prod_apply ~> rel_prod_inject
      pred_prod_apply ~> pred_prod_inject
    INCOMPATIBILITY.
  - The "size" plugin has been made compatible again with locales.
  - The theorems about "rel" and "set" may have a slightly different (but
    equivalent) form.
    INCOMPATIBILITY.

* The 'coinductive' command produces a proper coinduction rule for
mutual coinductive predicates. This new rule replaces the old rule,
which exposed details of the internal fixpoint construction and was
hard to use. INCOMPATIBILITY.

* New abbreviations for negated existence (but not bounded existence):

  \<nexists>x. P x \<equiv> \<not> (\<exists>x. P x)
  \<nexists>!x. P x \<equiv> \<not> (\<exists>!x. P x)

* The print mode "HOL" for ASCII syntax of binders "!", "?", "?!", "@"
has been removed for output. It is retained for input only, until it is
eliminated altogether.

* The unique existence quantifier no longer provides 'binder' syntax,
but uses syntax translations (as for bounded unique existence). Thus
iterated quantification \<exists>!x y. P x y with its slightly confusing
sequential meaning \<exists>!x. \<exists>!y. P x y is no longer possible. Instead,
pattern abstraction admits simultaneous unique existence \<exists>!(x, y). P x y
(analogous to existing notation \<exists>!(x, y)\<in>A. P x y). Potential
INCOMPATIBILITY in rare situations.

* Conventional syntax "%(). t" for unit abstractions. Slight syntactic
INCOMPATIBILITY.

* Renamed constants and corresponding theorems:

    setsum ~> sum
    setprod ~> prod
    listsum ~> sum_list
    listprod ~> prod_list

INCOMPATIBILITY.

* Sligthly more standardized theorem names:
    sgn_times ~> sgn_mult
    sgn_mult' ~> Real_Vector_Spaces.sgn_mult
    divide_zero_left ~> div_0
    zero_mod_left ~> mod_0
    divide_zero ~> div_by_0
    divide_1 ~> div_by_1
    nonzero_mult_divide_cancel_left ~> nonzero_mult_div_cancel_left
    div_mult_self1_is_id ~> nonzero_mult_div_cancel_left
    nonzero_mult_divide_cancel_right ~> nonzero_mult_div_cancel_right
    div_mult_self2_is_id ~> nonzero_mult_div_cancel_right
    is_unit_divide_mult_cancel_left ~> is_unit_div_mult_cancel_left
    is_unit_divide_mult_cancel_right ~> is_unit_div_mult_cancel_right
    mod_div_equality ~> div_mult_mod_eq
    mod_div_equality2 ~> mult_div_mod_eq
    mod_div_equality3 ~> mod_div_mult_eq
    mod_div_equality4 ~> mod_mult_div_eq
    minus_div_eq_mod ~> minus_div_mult_eq_mod
    minus_div_eq_mod2 ~> minus_mult_div_eq_mod
    minus_mod_eq_div ~> minus_mod_eq_div_mult
    minus_mod_eq_div2 ~> minus_mod_eq_mult_div
    div_mod_equality' ~> minus_mod_eq_div_mult [symmetric]
    mod_div_equality' ~> minus_div_mult_eq_mod [symmetric]
    zmod_zdiv_equality ~> mult_div_mod_eq [symmetric]
    zmod_zdiv_equality' ~> minus_div_mult_eq_mod [symmetric]
    Divides.mult_div_cancel ~> minus_mod_eq_mult_div [symmetric]
    mult_div_cancel ~> minus_mod_eq_mult_div [symmetric]
    zmult_div_cancel ~> minus_mod_eq_mult_div [symmetric]
    div_1 ~> div_by_Suc_0
    mod_1 ~> mod_by_Suc_0
INCOMPATIBILITY.

* New type class "idom_abs_sgn" specifies algebraic properties
of sign and absolute value functions.  Type class "sgn_if" has
disappeared.  Slight INCOMPATIBILITY.

* Dedicated syntax LENGTH('a) for length of types.

* Characters (type char) are modelled as finite algebraic type
corresponding to {0..255}.

  - Logical representation:
    * 0 is instantiated to the ASCII zero character.
    * All other characters are represented as "Char n"
      with n being a raw numeral expression less than 256.
    * Expressions of the form "Char n" with n greater than 255
      are non-canonical.
  - Printing and parsing:
    * Printable characters are printed and parsed as "CHR ''\<dots>''"
      (as before).
    * The ASCII zero character is printed and parsed as "0".
    * All other canonical characters are printed as "CHR 0xXX"
      with XX being the hexadecimal character code.  "CHR n"
      is parsable for every numeral expression n.
    * Non-canonical characters have no special syntax and are
      printed as their logical representation.
  - Explicit conversions from and to the natural numbers are
    provided as char_of_nat, nat_of_char (as before).
  - The auxiliary nibble type has been discontinued.

INCOMPATIBILITY.

* Type class "div" with operation "mod" renamed to type class "modulo"
with operation "modulo", analogously to type class "divide". This
eliminates the need to qualify any of those names in the presence of
infix "mod" syntax. INCOMPATIBILITY.

* Statements and proofs of Knaster-Tarski fixpoint combinators lfp/gfp
have been clarified. The fixpoint properties are lfp_fixpoint, its
symmetric lfp_unfold (as before), and the duals for gfp. Auxiliary items
for the proof (lfp_lemma2 etc.) are no longer exported, but can be
easily recovered by composition with eq_refl. Minor INCOMPATIBILITY.

* Constant "surj" is a mere input abbreviation, to avoid hiding an
equation in term output. Minor INCOMPATIBILITY.

* Command 'code_reflect' accepts empty constructor lists for datatypes,
which renders those abstract effectively.

* Command 'export_code' checks given constants for abstraction
violations: a small guarantee that given constants specify a safe
interface for the generated code.

* Code generation for Scala: ambiguous implicts in class diagrams are
spelt out explicitly.

* Static evaluators (Code_Evaluation.static_* in Isabelle/ML) rely on
explicitly provided auxiliary definitions for required type class
dictionaries rather than half-working magic. INCOMPATIBILITY, see the
tutorial on code generation for details.

* Theory Set_Interval: substantial new theorems on indexed sums and
products.

* Locale bijection establishes convenient default simp rules such as
"inv f (f a) = a" for total bijections.

* Abstract locales semigroup, abel_semigroup, semilattice,
semilattice_neutr, ordering, ordering_top, semilattice_order,
semilattice_neutr_order, comm_monoid_set, semilattice_set,
semilattice_neutr_set, semilattice_order_set,
semilattice_order_neutr_set monoid_list, comm_monoid_list,
comm_monoid_list_set, comm_monoid_mset, comm_monoid_fun use boldified
syntax uniformly that does not clash with corresponding global syntax.
INCOMPATIBILITY.

* Former locale lifting_syntax is now a bundle, which is easier to
include in a local context or theorem statement, e.g. "context includes
lifting_syntax begin ... end". Minor INCOMPATIBILITY.

* Some old / obsolete theorems have been renamed / removed, potential
INCOMPATIBILITY.

  nat_less_cases  --  removed, use linorder_cases instead
  inv_image_comp  --  removed, use image_inv_f_f instead
  image_surj_f_inv_f  ~>  image_f_inv_f

* Some theorems about groups and orders have been generalised from
  groups to semi-groups that are also monoids:
    le_add_same_cancel1
    le_add_same_cancel2
    less_add_same_cancel1
    less_add_same_cancel2
    add_le_same_cancel1
    add_le_same_cancel2
    add_less_same_cancel1
    add_less_same_cancel2

* Some simplifications theorems about rings have been removed, since
  superseeded by a more general version:
    less_add_cancel_left_greater_zero ~> less_add_same_cancel1
    less_add_cancel_right_greater_zero ~> less_add_same_cancel2
    less_eq_add_cancel_left_greater_eq_zero ~> le_add_same_cancel1
    less_eq_add_cancel_right_greater_eq_zero ~> le_add_same_cancel2
    less_eq_add_cancel_left_less_eq_zero ~> add_le_same_cancel1
    less_eq_add_cancel_right_less_eq_zero ~> add_le_same_cancel2
    less_add_cancel_left_less_zero ~> add_less_same_cancel1
    less_add_cancel_right_less_zero ~> add_less_same_cancel2
INCOMPATIBILITY.

* Renamed split_if -> if_split and split_if_asm -> if_split_asm to
resemble the f.split naming convention, INCOMPATIBILITY.

* Added class topological_monoid.

* The following theorems have been renamed:

  setsum_left_distrib ~> sum_distrib_right
  setsum_right_distrib ~> sum_distrib_left

INCOMPATIBILITY.

* Compound constants INFIMUM and SUPREMUM are mere abbreviations now.
INCOMPATIBILITY.

* "Gcd (f ` A)" and "Lcm (f ` A)" are printed with optional
comprehension-like syntax analogously to "Inf (f ` A)" and "Sup (f `
A)".

* Class semiring_Lcd merged into semiring_Gcd. INCOMPATIBILITY.

* The type class ordered_comm_monoid_add is now called
ordered_cancel_comm_monoid_add. A new type class ordered_comm_monoid_add
is introduced as the combination of ordered_ab_semigroup_add +
comm_monoid_add. INCOMPATIBILITY.

* Introduced the type classes canonically_ordered_comm_monoid_add and
dioid.

* Introduced the type class ordered_ab_semigroup_monoid_add_imp_le. When
instantiating linordered_semiring_strict and ordered_ab_group_add, an
explicit instantiation of ordered_ab_semigroup_monoid_add_imp_le might
be required. INCOMPATIBILITY.

* Dropped various legacy fact bindings, whose replacements are often
of a more general type also:
  lcm_left_commute_nat ~> lcm.left_commute
  lcm_left_commute_int ~> lcm.left_commute
  gcd_left_commute_nat ~> gcd.left_commute
  gcd_left_commute_int ~> gcd.left_commute
  gcd_greatest_iff_nat ~> gcd_greatest_iff
  gcd_greatest_iff_int ~> gcd_greatest_iff
  coprime_dvd_mult_nat ~> coprime_dvd_mult
  coprime_dvd_mult_int ~> coprime_dvd_mult
  zpower_numeral_even ~> power_numeral_even
  gcd_mult_cancel_nat ~> gcd_mult_cancel
  gcd_mult_cancel_int ~> gcd_mult_cancel
  div_gcd_coprime_nat ~> div_gcd_coprime
  div_gcd_coprime_int ~> div_gcd_coprime
  zpower_numeral_odd ~> power_numeral_odd
  zero_less_int_conv ~> of_nat_0_less_iff
  gcd_greatest_nat ~> gcd_greatest
  gcd_greatest_int ~> gcd_greatest
  coprime_mult_nat ~> coprime_mult
  coprime_mult_int ~> coprime_mult
  lcm_commute_nat ~> lcm.commute
  lcm_commute_int ~> lcm.commute
  int_less_0_conv ~> of_nat_less_0_iff
  gcd_commute_nat ~> gcd.commute
  gcd_commute_int ~> gcd.commute
  Gcd_insert_nat ~> Gcd_insert
  Gcd_insert_int ~> Gcd_insert
  of_int_int_eq ~> of_int_of_nat_eq
  lcm_least_nat ~> lcm_least
  lcm_least_int ~> lcm_least
  lcm_assoc_nat ~> lcm.assoc
  lcm_assoc_int ~> lcm.assoc
  int_le_0_conv ~> of_nat_le_0_iff
  int_eq_0_conv ~> of_nat_eq_0_iff
  Gcd_empty_nat ~> Gcd_empty
  Gcd_empty_int ~> Gcd_empty
  gcd_assoc_nat ~> gcd.assoc
  gcd_assoc_int ~> gcd.assoc
  zero_zle_int ~> of_nat_0_le_iff
  lcm_dvd2_nat ~> dvd_lcm2
  lcm_dvd2_int ~> dvd_lcm2
  lcm_dvd1_nat ~> dvd_lcm1
  lcm_dvd1_int ~> dvd_lcm1
  gcd_zero_nat ~> gcd_eq_0_iff
  gcd_zero_int ~> gcd_eq_0_iff
  gcd_dvd2_nat ~> gcd_dvd2
  gcd_dvd2_int ~> gcd_dvd2
  gcd_dvd1_nat ~> gcd_dvd1
  gcd_dvd1_int ~> gcd_dvd1
  int_numeral ~> of_nat_numeral
  lcm_ac_nat ~> ac_simps
  lcm_ac_int ~> ac_simps
  gcd_ac_nat ~> ac_simps
  gcd_ac_int ~> ac_simps
  abs_int_eq ~> abs_of_nat
  zless_int ~> of_nat_less_iff
  zdiff_int ~> of_nat_diff
  zadd_int ~> of_nat_add
  int_mult ~> of_nat_mult
  int_Suc ~> of_nat_Suc
  inj_int ~> inj_of_nat
  int_1 ~> of_nat_1
  int_0 ~> of_nat_0
  Lcm_empty_nat ~> Lcm_empty
  Lcm_empty_int ~> Lcm_empty
  Lcm_insert_nat ~> Lcm_insert
  Lcm_insert_int ~> Lcm_insert
  comp_fun_idem_gcd_nat ~> comp_fun_idem_gcd
  comp_fun_idem_gcd_int ~> comp_fun_idem_gcd
  comp_fun_idem_lcm_nat ~> comp_fun_idem_lcm