NEWS

provide @{file_unchecked} in Isabelle/Pure;

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

New in this Isabelle version
----------------------------

*** General ***

* Document antiquotation @{url} produces markup for the given URL,
which results in an active hyperlink within the text.

* Document antiquotation @{file_unchecked} is like @{file}, but does
not check existence within the file-system.


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

* Auxiliary files ('ML_file' etc.) are managed by the Prover IDE.
Open text buffers take precedence over copies within the file-system.

* Document panel: simplied interaction where every single mouse click
(re)opens document via desktop environment or as jEdit buffer.


*** HOL ***

* Qualified constant names Wellfounded.acc, Wellfounded.accp.
INCOMPATIBILITY.

* Fact generalization and consolidation:
    neq_one_mod_two, mod_2_not_eq_zero_eq_one_int ~> not_mod_2_eq_0_eq_1
INCOMPATIBILITY.

* Purely algebraic definition of even.  Fact generalization and consolidation:
    nat_even_iff_2_dvd, int_even_iff_2_dvd ~> even_iff_2_dvd
    even_zero_(nat|int) ~> even_zero
INCOMPATIBILITY.

* Abolished neg_numeral.
  * Canonical representation for minus one is "- 1".
  * Canonical representation for other negative numbers is "- (numeral _)".
  * When devising rule sets for number calculation, consider the
    following canonical cases: 0, 1, numeral _, - 1, - numeral _.
  * HOLogic.dest_number also recognizes numerals in non-canonical forms
    like "numeral One", "- numeral One", "- 0" and even "- … - _".
  * Syntax for negative numerals is mere input syntax.
INCOMPATBILITY.

* Elimination of fact duplicates:
    equals_zero_I ~> minus_unique
    diff_eq_0_iff_eq ~> right_minus_eq
    nat_infinite ~> infinite_UNIV_nat
    int_infinite ~> infinite_UNIV_int
INCOMPATIBILITY.

* Fact name consolidation:
    diff_def, diff_minus, ab_diff_minus ~> diff_conv_add_uminus
    minus_le_self_iff ~> neg_less_eq_nonneg
    le_minus_self_iff ~> less_eq_neg_nonpos
    neg_less_nonneg ~> neg_less_pos
    less_minus_self_iff ~> less_neg_neg [simp]
INCOMPATIBILITY.

* More simplification rules on unary and binary minus:
add_diff_cancel, add_diff_cancel_left, add_le_same_cancel1,
add_le_same_cancel2, add_less_same_cancel1, add_less_same_cancel2,
add_minus_cancel, diff_add_cancel, le_add_same_cancel1,
le_add_same_cancel2, less_add_same_cancel1, less_add_same_cancel2,
minus_add_cancel, uminus_add_conv_diff.  These correspondingly
have been taken away from fact collections algebra_simps and
field_simps.  INCOMPATIBILITY.

To restore proofs, the following patterns are helpful:

a) Arbitrary failing proof not involving "diff_def":
Consider simplification with algebra_simps or field_simps.

b) Lifting rules from addition to subtraction:
Try with "using <rule for addition> of [… "- _" …]" by simp".

c) Simplification with "diff_def": just drop "diff_def".
Consider simplification with algebra_simps or field_simps;
or the brute way with
"simp add: diff_conv_add_uminus del: add_uminus_conv_diff".

* SUP and INF generalized to conditionally_complete_lattice

* Theory Lubs moved HOL image to HOL-Library. It is replaced by
Conditionally_Complete_Lattices.   INCOMPATIBILITY.

* Introduce bdd_above and bdd_below in Conditionally_Complete_Lattices, use them
instead of explicitly stating boundedness of sets.

* ccpo.admissible quantifies only over non-empty chains to allow
more syntax-directed proof rules; the case of the empty chain
shows up as additional case in fixpoint induction proofs.
INCOMPATIBILITY

* Nitpick:
  - Fixed soundness bug whereby mutually recursive datatypes could take
    infinite values.


*** ML ***

* Toplevel function "use" refers to raw ML bootstrap environment,
without Isar context nor antiquotations.  Potential INCOMPATIBILITY.
Note that 'ML_file' is the canonical command to load ML files into the
formal context.


*** System ***

* Simplified "isabelle display" tool.  Settings variables DVI_VIEWER
and PDF_VIEWER now refer to the actual programs, not shell
command-lines.  Discontinued option -c: invocation may be asynchronous
via desktop environment, without any special precautions.  Potential
INCOMPATIBILITY with ambitious private settings.

* Improved 'display_drafts' concerning desktop integration and
repeated invocation in PIDE front-end: re-use single file
$ISABELLE_HOME_USER/tmp/drafts.pdf and corresponding views.



New in Isabelle2013-2 (December 2013)
-------------------------------------

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

* More robust editing of running commands with internal forks,
e.g. non-terminating 'by' steps.

* More relaxed Sledgehammer panel: avoid repeated application of query
after edits surrounding the command location.

* More status information about commands that are interrupted
accidentally (via physical event or Poly/ML runtime system signal,
e.g. out-of-memory).


*** System ***

* More robust termination of external processes managed by
Isabelle/ML: support cancellation of tasks within the range of
milliseconds, as required for PIDE document editing with automatically
tried tools (e.g. Sledgehammer).

* Reactivated Isabelle/Scala kill command for external processes on
Mac OS X, which was accidentally broken in Isabelle2013-1 due to a
workaround for some Debian/Ubuntu Linux versions from 2013.



New in Isabelle2013-1 (November 2013)
-------------------------------------

*** General ***

* Discontinued obsolete 'uses' within theory header.  Note that
commands like 'ML_file' work without separate declaration of file
dependencies.  Minor INCOMPATIBILITY.

* Discontinued redundant 'use' command, which was superseded by
'ML_file' in Isabelle2013.  Minor INCOMPATIBILITY.

* Simplified subscripts within identifiers, using plain \<^sub>
instead of the second copy \<^isub> and \<^isup>.  Superscripts are
only for literal tokens within notation; explicit mixfix annotations
for consts or fixed variables may be used as fall-back for unusual
names.  Obsolete \<twosuperior> has been expanded to \<^sup>2 in
Isabelle/HOL.  INCOMPATIBILITY, use "isabelle update_sub_sup" to
standardize symbols as a starting point for further manual cleanup.
The ML reference variable "legacy_isub_isup" may be set as temporary
workaround, to make the prover accept a subset of the old identifier
syntax.

* Document antiquotations: term style "isub" has been renamed to
"sub".  Minor INCOMPATIBILITY.

* Uniform management of "quick_and_dirty" as system option (see also
"isabelle options"), configuration option within the context (see also
Config.get in Isabelle/ML), and attribute in Isabelle/Isar.  Minor
INCOMPATIBILITY, need to use more official Isabelle means to access
quick_and_dirty, instead of historical poking into mutable reference.

* Renamed command 'print_configs' to 'print_options'.  Minor
INCOMPATIBILITY.

* Proper diagnostic command 'print_state'.  Old 'pr' (with its
implicit change of some global references) is retained for now as
control command, e.g. for ProofGeneral 3.7.x.

* Discontinued 'print_drafts' command with its old-fashioned PS output
and Unix command-line print spooling.  Minor INCOMPATIBILITY: use
'display_drafts' instead and print via the regular document viewer.

* Updated and extended "isar-ref" and "implementation" manual,
eliminated old "ref" manual.


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

* New manual "jedit" for Isabelle/jEdit, see isabelle doc or
Documentation panel.

* Dockable window "Documentation" provides access to Isabelle
documentation.

* Dockable window "Find" provides query operations for formal entities
(GUI front-end to 'find_theorems' command).

* Dockable window "Sledgehammer" manages asynchronous / parallel
sledgehammer runs over existing document sources, independently of
normal editing and checking process.

* Dockable window "Timing" provides an overview of relevant command
timing information, depending on option jedit_timing_threshold.  The
same timing information is shown in the extended tooltip of the
command keyword, when hovering the mouse over it while the CONTROL or
COMMAND modifier is pressed.

* Improved dockable window "Theories": Continuous checking of proof
document (visible and required parts) may be controlled explicitly,
using check box or shortcut "C+e ENTER".  Individual theory nodes may
be marked explicitly as required and checked in full, using check box
or shortcut "C+e SPACE".

* Improved completion mechanism, which is now managed by the
Isabelle/jEdit plugin instead of SideKick.  Refined table of Isabelle
symbol abbreviations (see $ISABELLE_HOME/etc/symbols).

* Standard jEdit keyboard shortcut C+b complete-word is remapped to
isabelle.complete for explicit completion in Isabelle sources.
INCOMPATIBILITY wrt. jEdit defaults, may have to invent new shortcuts
to resolve conflict.

* Improved support of various "minor modes" for Isabelle NEWS,
options, session ROOT etc., with completion and SideKick tree view.

* Strictly monotonic document update, without premature cancellation of
running transactions that are still needed: avoid reset/restart of
such command executions while editing.

* Support for asynchronous print functions, as overlay to existing
document content.

* Support for automatic tools in HOL, which try to prove or disprove
toplevel theorem statements.

* Action isabelle.reset-font-size resets main text area font size
according to Isabelle/Scala plugin option "jedit_font_reset_size" (see
also "Plugin Options / Isabelle / General").  It can be bound to some
keyboard shortcut by the user (e.g. C+0 and/or C+NUMPAD0).

* File specifications in jEdit (e.g. file browser) may refer to
$ISABELLE_HOME and $ISABELLE_HOME_USER on all platforms.  Discontinued
obsolete $ISABELLE_HOME_WINDOWS variable.

* Improved support for Linux look-and-feel "GTK+", see also "Utilities
/ Global Options / Appearance".

* Improved support of native Mac OS X functionality via "MacOSX"
plugin, which is now enabled by default.


*** Pure ***

* Commands 'interpretation' and 'sublocale' are now target-sensitive.
In particular, 'interpretation' allows for non-persistent
interpretation within "context ... begin ... end" blocks offering a
light-weight alternative to 'sublocale'.  See "isar-ref" manual for
details.

* Improved locales diagnostic command 'print_dependencies'.

* Discontinued obsolete 'axioms' command, which has been marked as
legacy since Isabelle2009-2.  INCOMPATIBILITY, use 'axiomatization'
instead, while observing its uniform scope for polymorphism.

* Discontinued empty name bindings in 'axiomatization'.
INCOMPATIBILITY.

* System option "proofs" has been discontinued.  Instead the global
state of Proofterm.proofs is persistently compiled into logic images
as required, notably HOL-Proofs.  Users no longer need to change
Proofterm.proofs dynamically.  Minor INCOMPATIBILITY.

* Syntax translation functions (print_translation etc.) always depend
on Proof.context.  Discontinued former "(advanced)" option -- this is
now the default.  Minor INCOMPATIBILITY.

* Former global reference trace_unify_fail is now available as
configuration option "unify_trace_failure" (global context only).

* SELECT_GOAL now retains the syntactic context of the overall goal
state (schematic variables etc.).  Potential INCOMPATIBILITY in rare
situations.


*** HOL ***

* Stronger precedence of syntax for big intersection and union on
sets, in accordance with corresponding lattice operations.
INCOMPATIBILITY.

* Notation "{p:A. P}" now allows tuple patterns as well.

* Nested case expressions are now translated in a separate check phase
rather than during parsing. The data for case combinators is separated
from the datatype package. The declaration attribute
"case_translation" can be used to register new case combinators:

  declare [[case_translation case_combinator constructor1 ... constructorN]]

* Code generator:
  - 'code_printing' unifies 'code_const' / 'code_type' / 'code_class' /
    'code_instance'.
  - 'code_identifier' declares name hints for arbitrary identifiers in
    generated code, subsuming 'code_modulename'.

See the isar-ref manual for syntax diagrams, and the HOL theories for
examples.

* Attibute 'code': 'code' now declares concrete and abstract code
equations uniformly.  Use explicit 'code equation' and 'code abstract'
to distinguish both when desired.

* Discontinued theories Code_Integer and Efficient_Nat by a more
fine-grain stack of theories Code_Target_Int, Code_Binary_Nat,
Code_Target_Nat and Code_Target_Numeral.  See the tutorial on code
generation for details.  INCOMPATIBILITY.

* Numeric types are mapped by default to target language numerals:
natural (replaces former code_numeral) and integer (replaces former
code_int).  Conversions are available as integer_of_natural /
natural_of_integer / integer_of_nat / nat_of_integer (in HOL) and
Code_Numeral.integer_of_natural / Code_Numeral.natural_of_integer (in
ML).  INCOMPATIBILITY.

* Function package: For mutually recursive functions f and g, separate
cases rules f.cases and g.cases are generated instead of unusable
f_g.cases which exposed internal sum types. Potential INCOMPATIBILITY,
in the case that the unusable rule was used nevertheless.

* Function package: For each function f, new rules f.elims are
generated, which eliminate equalities of the form "f x = t".

* New command 'fun_cases' derives ad-hoc elimination rules for
function equations as simplified instances of f.elims, analogous to
inductive_cases.  See ~~/src/HOL/ex/Fundefs.thy for some examples.

* Lifting:
  - parametrized correspondence relations are now supported:
    + parametricity theorems for the raw term can be specified in
      the command lift_definition, which allow us to generate stronger
      transfer rules
    + setup_lifting generates stronger transfer rules if parametric
      correspondence relation can be generated
    + various new properties of the relator must be specified to support
      parametricity
    + parametricity theorem for the Quotient relation can be specified
  - setup_lifting generates domain rules for the Transfer package
  - stronger reflexivity prover of respectfulness theorems for type
    copies
  - ===> and --> are now local. The symbols can be introduced
    by interpreting the locale lifting_syntax (typically in an
    anonymous context)
  - Lifting/Transfer relevant parts of Library/Quotient_* are now in
    Main. Potential INCOMPATIBILITY
  - new commands for restoring and deleting Lifting/Transfer context:
    lifting_forget, lifting_update
  - the command print_quotmaps was renamed to print_quot_maps.
    INCOMPATIBILITY

* Transfer:
  - better support for domains in Transfer: replace Domainp T
    by the actual invariant in a transferred goal
  - transfer rules can have as assumptions other transfer rules
  - Experimental support for transferring from the raw level to the
    abstract level: Transfer.transferred attribute
  - Attribute version of the transfer method: untransferred attribute

* Reification and reflection:
  - Reification is now directly available in HOL-Main in structure
    "Reification".
  - Reflection now handles multiple lists with variables also.
  - The whole reflection stack has been decomposed into conversions.
INCOMPATIBILITY.

* Revised devices for recursive definitions over finite sets:
  - Only one fundamental fold combinator on finite set remains:
    Finite_Set.fold :: ('a => 'b => 'b) => 'b => 'a set => 'b
    This is now identity on infinite sets.
  - Locales ("mini packages") for fundamental definitions with
    Finite_Set.fold: folding, folding_idem.
  - Locales comm_monoid_set, semilattice_order_set and
    semilattice_neutr_order_set for big operators on sets.
    See theory Big_Operators for canonical examples.
    Note that foundational constants comm_monoid_set.F and
    semilattice_set.F correspond to former combinators fold_image
    and fold1 respectively.  These are now gone.  You may use
    those foundational constants as substitutes, but it is
    preferable to interpret the above locales accordingly.
  - Dropped class ab_semigroup_idem_mult (special case of lattice,
    no longer needed in connection with Finite_Set.fold etc.)
  - Fact renames:
      card.union_inter ~> card_Un_Int [symmetric]
      card.union_disjoint ~> card_Un_disjoint
INCOMPATIBILITY.

* Locale hierarchy for abstract orderings and (semi)lattices.

* Complete_Partial_Order.admissible is defined outside the type class
ccpo, but with mandatory prefix ccpo. Admissibility theorems lose the
class predicate assumption or sort constraint when possible.
INCOMPATIBILITY.

* Introduce type class "conditionally_complete_lattice": Like a
complete lattice but does not assume the existence of the top and
bottom elements.  Allows to generalize some lemmas about reals and
extended reals.  Removed SupInf and replaced it by the instantiation
of conditionally_complete_lattice for real. Renamed lemmas about
conditionally-complete lattice from Sup_... to cSup_... and from
Inf_...  to cInf_... to avoid hidding of similar complete lattice
lemmas.

* Introduce type class linear_continuum as combination of
conditionally-complete lattices and inner dense linorders which have
more than one element.  INCOMPATIBILITY.

* Introduced type classes order_top and order_bot. The old classes top
and bot only contain the syntax without assumptions.  INCOMPATIBILITY:
Rename bot -> order_bot, top -> order_top

* Introduce type classes "no_top" and "no_bot" for orderings without
top and bottom elements.

* Split dense_linorder into inner_dense_order and no_top, no_bot.

* Complex_Main: Unify and move various concepts from
HOL-Multivariate_Analysis to HOL-Complex_Main.

 - Introduce type class (lin)order_topology and
   linear_continuum_topology.  Allows to generalize theorems about
   limits and order.  Instances are reals and extended reals.

 - continuous and continuos_on from Multivariate_Analysis:
   "continuous" is the continuity of a function at a filter.  "isCont"
   is now an abbrevitation: "isCont x f == continuous (at _) f".

   Generalized continuity lemmas from isCont to continuous on an
   arbitrary filter.

 - compact from Multivariate_Analysis. Use Bolzano's lemma to prove
   compactness of closed intervals on reals. Continuous functions
   attain infimum and supremum on compact sets. The inverse of a
   continuous function is continuous, when the function is continuous
   on a compact set.

 - connected from Multivariate_Analysis. Use it to prove the
   intermediate value theorem. Show connectedness of intervals on
   linear_continuum_topology).

 - first_countable_topology from Multivariate_Analysis. Is used to
   show equivalence of properties on the neighbourhood filter of x and
   on all sequences converging to x.

 - FDERIV: Definition of has_derivative moved to Deriv.thy. Moved
   theorems from Library/FDERIV.thy to Deriv.thy and base the
   definition of DERIV on FDERIV. Add variants of DERIV and FDERIV
   which are restricted to sets, i.e. to represent derivatives from
   left or right.

 - Removed the within-filter. It is replaced by the principal filter:

     F within X = inf F (principal X)

 - Introduce "at x within U" as a single constant, "at x" is now an
   abbreviation for "at x within UNIV"

 - Introduce named theorem collections tendsto_intros,
   continuous_intros, continuous_on_intros and FDERIV_intros. Theorems
   in tendsto_intros (or FDERIV_intros) are also available as
   tendsto_eq_intros (or FDERIV_eq_intros) where the right-hand side
   is replaced by a congruence rule. This allows to apply them as
   intro rules and then proving equivalence by the simplifier.

 - Restructured theories in HOL-Complex_Main:

   + Moved RealDef and RComplete into Real

   + Introduced Topological_Spaces and moved theorems about
     topological spaces, filters, limits and continuity to it

   + Renamed RealVector to Real_Vector_Spaces

   + Split Lim, SEQ, Series into Topological_Spaces,
     Real_Vector_Spaces, and Limits

   + Moved Ln and Log to Transcendental

   + Moved theorems about continuity from Deriv to Topological_Spaces

 - Remove various auxiliary lemmas.

INCOMPATIBILITY.

* Nitpick:
  - Added option "spy"
  - Reduce incidence of "too high arity" errors

* Sledgehammer:
  - Renamed option:
      isar_shrink ~> isar_compress
    INCOMPATIBILITY.
  - Added options "isar_try0", "spy"
  - Better support for "isar_proofs"
  - MaSh has been fined-tuned and now runs as a local server

* Improved support for ad hoc overloading of constants (see also
isar-ref manual and ~~/src/HOL/ex/Adhoc_Overloading_Examples.thy).

* Library/Polynomial.thy:
  - Use lifting for primitive definitions.
  - Explicit conversions from and to lists of coefficients, used for
    generated code.
  - Replaced recursion operator poly_rec by fold_coeffs.
  - Prefer pre-existing gcd operation for gcd.
  - Fact renames:
    poly_eq_iff ~> poly_eq_poly_eq_iff
    poly_ext ~> poly_eqI
    expand_poly_eq ~> poly_eq_iff
IMCOMPATIBILITY.

* New Library/Simps_Case_Conv.thy: Provides commands simps_of_case and
case_of_simps to convert function definitions between a list of
equations with patterns on the lhs and a single equation with case
expressions on the rhs. See also Ex/Simps_Case_Conv_Examples.thy.

* New Library/FSet.thy: type of finite sets defined as a subtype of
sets defined by Lifting/Transfer.

* Discontinued theory src/HOL/Library/Eval_Witness.  INCOMPATIBILITY.

* Consolidation of library theories on product orders:

    Product_Lattice ~> Product_Order -- pointwise order on products
    Product_ord ~> Product_Lexorder -- lexicographic order on products

INCOMPATIBILITY.

* Imperative-HOL: The MREC combinator is considered legacy and no
longer included by default. INCOMPATIBILITY, use partial_function
instead, or import theory Legacy_Mrec as a fallback.

* HOL-Algebra: Discontinued theories ~~/src/HOL/Algebra/abstract and
~~/src/HOL/Algebra/poly.  Existing theories should be based on
~~/src/HOL/Library/Polynomial instead.  The latter provides
integration with HOL's type classes for rings.  INCOMPATIBILITY.

* HOL-BNF:
  - Various improvements to BNF-based (co)datatype package, including
    new commands "primrec_new", "primcorec", and
    "datatype_new_compat", as well as documentation. See
    "datatypes.pdf" for details.
  - New "coinduction" method to avoid some boilerplate (compared to
    coinduct).
  - Renamed keywords:
    data ~> datatype_new
    codata ~> codatatype
    bnf_def ~> bnf
  - Renamed many generated theorems, including
    discs ~> disc
    map_comp' ~> map_comp
    map_id' ~> map_id
    sels ~> sel
    set_map' ~> set_map
    sets ~> set
IMCOMPATIBILITY.


*** ML ***

* Spec_Check is a Quickcheck tool for Isabelle/ML.  The ML function
"check_property" allows to check specifications of the form "ALL x y
z. prop x y z".  See also ~~/src/Tools/Spec_Check/ with its
Examples.thy in particular.

* Improved printing of exception trace in Poly/ML 5.5.1, with regular
tracing output in the command transaction context instead of physical
stdout.  See also Toplevel.debug, Toplevel.debugging and
ML_Compiler.exn_trace.

* ML type "theory" is now immutable, without any special treatment of
drafts or linear updates (which could lead to "stale theory" errors in
the past).  Discontinued obsolete operations like Theory.copy,
Theory.checkpoint, and the auxiliary type theory_ref.  Minor
INCOMPATIBILITY.

* More uniform naming of goal functions for skipped proofs:

    Skip_Proof.prove  ~>  Goal.prove_sorry
    Skip_Proof.prove_global  ~>  Goal.prove_sorry_global

Minor INCOMPATIBILITY.

* Simplifier tactics and tools use proper Proof.context instead of
historic type simpset.  Old-style declarations like addsimps,
addsimprocs etc. operate directly on Proof.context.  Raw type simpset
retains its use as snapshot of the main Simplifier context, using
simpset_of and put_simpset on Proof.context.  INCOMPATIBILITY -- port
old tools by making them depend on (ctxt : Proof.context) instead of
(ss : simpset), then turn (simpset_of ctxt) into ctxt.

* Modifiers for classical wrappers (e.g. addWrapper, delWrapper)
operate on Proof.context instead of claset, for uniformity with addIs,
addEs, addDs etc. Note that claset_of and put_claset allow to manage
clasets separately from the context.

* Discontinued obsolete ML antiquotations @{claset} and @{simpset}.
INCOMPATIBILITY, use @{context} instead.

* Antiquotation @{theory_context A} is similar to @{theory A}, but
presents the result as initial Proof.context.


*** System ***

* Discontinued obsolete isabelle usedir, mkdir, make -- superseded by
"isabelle build" in Isabelle2013.  INCOMPATIBILITY.

* Discontinued obsolete isabelle-process options -f and -u (former
administrative aliases of option -e).  Minor INCOMPATIBILITY.

* Discontinued obsolete isabelle print tool, and PRINT_COMMAND
settings variable.

* Discontinued ISABELLE_DOC_FORMAT settings variable and historic
document formats: dvi.gz, ps, ps.gz -- the default document format is
always pdf.

* Isabelle settings variable ISABELLE_BUILD_JAVA_OPTIONS allows to
specify global resources of the JVM process run by isabelle build.

* Toplevel executable $ISABELLE_HOME/bin/isabelle_scala_script allows
to run Isabelle/Scala source files as standalone programs.

* Improved "isabelle keywords" tool (for old-style ProofGeneral
keyword tables): use Isabelle/Scala operations, which inspect outer
syntax without requiring to build sessions first.

* Sessions may be organized via 'chapter' specifications in the ROOT
file, which determines a two-level hierarchy of browser info.  The old
tree-like organization via implicit sub-session relation (with its
tendency towards erratic fluctuation of URLs) has been discontinued.
The default chapter is called "Unsorted".  Potential INCOMPATIBILITY
for HTML presentation of theories.



New in Isabelle2013 (February 2013)
-----------------------------------

*** General ***

* Theorem status about oracles and unfinished/failed future proofs is
no longer printed by default, since it is incompatible with
incremental / parallel checking of the persistent document model.  ML
function Thm.peek_status may be used to inspect a snapshot of the
ongoing evaluation process.  Note that in batch mode --- notably
isabelle build --- the system ensures that future proofs of all
accessible theorems in the theory context are finished (as before).

* Configuration option show_markup controls direct inlining of markup
into the printed representation of formal entities --- notably type
and sort constraints.  This enables Prover IDE users to retrieve that
information via tooltips in the output window, for example.

* Command 'ML_file' evaluates ML text from a file directly within the
theory, without any predeclaration via 'uses' in the theory header.

* Old command 'use' command and corresponding keyword 'uses' in the
theory header are legacy features and will be discontinued soon.
Tools that load their additional source files may imitate the
'ML_file' implementation, such that the system can take care of
dependencies properly.

* Discontinued obsolete method fastsimp / tactic fast_simp_tac, which
is called fastforce / fast_force_tac already since Isabelle2011-1.

* Updated and extended "isar-ref" and "implementation" manual, reduced
remaining material in old "ref" manual.

* Improved support for auxiliary contexts that indicate block structure
for specifications.  Nesting of "context fixes ... context assumes ..."
and "class ... context ...".

* Attribute "consumes" allows a negative value as well, which is
interpreted relatively to the total number of premises of the rule in
the target context.  This form of declaration is stable when exported
from a nested 'context' with additional assumptions.  It is the
preferred form for definitional packages, notably cases/rules produced
in HOL/inductive and HOL/function.

* More informative error messages for Isar proof commands involving
lazy enumerations (method applications etc.).

* Refined 'help' command to retrieve outer syntax commands according
to name patterns (with clickable results).


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

* Parallel terminal proofs ('by') are enabled by default, likewise
proofs that are built into packages like 'datatype', 'function'.  This
allows to "run ahead" checking the theory specifications on the
surface, while the prover is still crunching on internal
justifications.  Unfinished / cancelled proofs are restarted as
required to complete full proof checking eventually.

* Improved output panel with tooltips, hyperlinks etc. based on the
same Rich_Text_Area as regular Isabelle/jEdit buffers.  Activation of
tooltips leads to some window that supports the same recursively,
which can lead to stacks of tooltips as the semantic document content
is explored.  ESCAPE closes the whole stack, individual windows may be
closed separately, or detached to become independent jEdit dockables.

* Improved support for commands that produce graph output: the text
message contains a clickable area to open a new instance of the graph
browser on demand.

* More robust incremental parsing of outer syntax (partial comments,
malformed symbols).  Changing the balance of open/close quotes and
comment delimiters works more conveniently with unfinished situations
that frequently occur in user interaction.

* More efficient painting and improved reactivity when editing large
files.  More scalable management of formal document content.

* Smarter handling of tracing messages: prover process pauses after
certain number of messages per command transaction, with some user
dialog to stop or continue.  This avoids swamping the front-end with
potentially infinite message streams.

* More plugin options and preferences, based on Isabelle/Scala.  The
jEdit plugin option panel provides access to some Isabelle/Scala
options, including tuning parameters for editor reactivity and color
schemes.

* Dockable window "Symbols" provides some editing support for Isabelle
symbols.

* Dockable window "Monitor" shows ML runtime statistics.  Note that
continuous display of the chart slows down the system.

* Improved editing support for control styles: subscript, superscript,
bold, reset of style -- operating on single symbols or text
selections.  Cf. keyboard shortcuts C+e DOWN/UP/RIGHT/LEFT.

* Actions isabelle.increase-font-size and isabelle.decrease-font-size
adjust the main text area font size, and its derivatives for output,
tooltips etc.  Cf. keyboard shortcuts C-PLUS and C-MINUS, which often
need to be adapted to local keyboard layouts.

* More reactive completion popup by default: use \t (TAB) instead of
\n (NEWLINE) to minimize intrusion into regular flow of editing.  See
also "Plugin Options / SideKick / General / Code Completion Options".

* Implicit check and build dialog of the specified logic session
image.  For example, HOL, HOLCF, HOL-Nominal can be produced on
demand, without bundling big platform-dependent heap images in the
Isabelle distribution.

* Uniform Java 7 platform on Linux, Mac OS X, Windows: recent updates
from Oracle provide better multi-platform experience.  This version is
now bundled exclusively with Isabelle.


*** Pure ***

* Code generation for Haskell: restrict unqualified imports from
Haskell Prelude to a small set of fundamental operations.

* Command 'export_code': relative file names are interpreted
relatively to master directory of current theory rather than the
rather arbitrary current working directory.  INCOMPATIBILITY.

* Discontinued obsolete attribute "COMP".  Potential INCOMPATIBILITY,
use regular rule composition via "OF" / "THEN", or explicit proof
structure instead.  Note that Isabelle/ML provides a variety of
operators like COMP, INCR_COMP, COMP_INCR, which need to be applied
with some care where this is really required.

* Command 'typ' supports an additional variant with explicit sort
constraint, to infer and check the most general type conforming to a
given sort.  Example (in HOL):

  typ "_ * _ * bool * unit" :: finite

* Command 'locale_deps' visualizes all locales and their relations as
a Hasse diagram.


*** HOL ***

* Sledgehammer:

  - Added MaSh relevance filter based on machine-learning; see the
    Sledgehammer manual for details.
  - Polished Isar proofs generated with "isar_proofs" option.
  - Rationalized type encodings ("type_enc" option).
  - Renamed "kill_provers" subcommand to "kill_all".
  - Renamed options:
      isar_proof ~> isar_proofs
      isar_shrink_factor ~> isar_shrink
      max_relevant ~> max_facts
      relevance_thresholds ~> fact_thresholds

* Quickcheck: added an optimisation for equality premises.  It is
switched on by default, and can be switched off by setting the
configuration quickcheck_optimise_equality to false.

* Quotient: only one quotient can be defined by quotient_type
INCOMPATIBILITY.

* Lifting:
  - generation of an abstraction function equation in lift_definition
  - quot_del attribute
  - renamed no_abs_code -> no_code (INCOMPATIBILITY.)

* Simproc "finite_Collect" rewrites set comprehensions into pointfree
expressions.

* Preprocessing of the code generator rewrites set comprehensions into
pointfree expressions.

* The SMT solver Z3 has now by default a restricted set of directly
supported features. For the full set of features (div/mod, nonlinear
arithmetic, datatypes/records) with potential proof reconstruction
failures, enable the configuration option "z3_with_extensions".  Minor
INCOMPATIBILITY.

* Simplified 'typedef' specifications: historical options for implicit
set definition and alternative name have been discontinued.  The
former behavior of "typedef (open) t = A" is now the default, but
written just "typedef t = A".  INCOMPATIBILITY, need to adapt theories
accordingly.

* Removed constant "chars"; prefer "Enum.enum" on type "char"
directly.  INCOMPATIBILITY.

* Moved operation product, sublists and n_lists from theory Enum to
List.  INCOMPATIBILITY.

* Theorem UN_o generalized to SUP_comp.  INCOMPATIBILITY.

* Class "comm_monoid_diff" formalises properties of bounded
subtraction, with natural numbers and multisets as typical instances.

* Added combinator "Option.these" with type "'a option set => 'a set".

* Theory "Transitive_Closure": renamed lemmas

  reflcl_tranclp -> reflclp_tranclp
  rtranclp_reflcl -> rtranclp_reflclp

INCOMPATIBILITY.

* Theory "Rings": renamed lemmas (in class semiring)

  left_distrib ~> distrib_right
  right_distrib ~> distrib_left

INCOMPATIBILITY.

* Generalized the definition of limits:

  - Introduced the predicate filterlim (LIM x F. f x :> G) which
    expresses that when the input values x converge to F then the
    output f x converges to G.

  - Added filters for convergence to positive (at_top) and negative
    infinity (at_bot).

  - Moved infinity in the norm (at_infinity) from
    Multivariate_Analysis to Complex_Main.

  - Removed real_tendsto_inf, it is superseded by "LIM x F. f x :>
    at_top".

INCOMPATIBILITY.

* Theory "Library/Option_ord" provides instantiation of option type to
lattice type classes.

* Theory "Library/Multiset": renamed

    constant fold_mset ~> Multiset.fold
    fact fold_mset_commute ~> fold_mset_comm

INCOMPATIBILITY.

* Renamed theory Library/List_Prefix to Library/Sublist, with related
changes as follows.

  - Renamed constants (and related lemmas)

      prefix ~> prefixeq
      strict_prefix ~> prefix

  - Replaced constant "postfix" by "suffixeq" with swapped argument
    order (i.e., "postfix xs ys" is now "suffixeq ys xs") and dropped
    old infix syntax "xs >>= ys"; use "suffixeq ys xs" instead.
    Renamed lemmas accordingly.

  - Added constant "list_hembeq" for homeomorphic embedding on
    lists. Added abbreviation "sublisteq" for special case
    "list_hembeq (op =)".

  - Theory Library/Sublist no longer provides "order" and "bot" type
    class instances for the prefix order (merely corresponding locale
    interpretations). The type class instances are now in theory
    Library/Prefix_Order.

  - The sublist relation of theory Library/Sublist_Order is now based
    on "Sublist.sublisteq".  Renamed lemmas accordingly:

      le_list_append_le_same_iff ~> Sublist.sublisteq_append_le_same_iff
      le_list_append_mono ~> Sublist.list_hembeq_append_mono
      le_list_below_empty ~> Sublist.list_hembeq_Nil, Sublist.list_hembeq_Nil2
      le_list_Cons_EX ~> Sublist.list_hembeq_ConsD
      le_list_drop_Cons2 ~> Sublist.sublisteq_Cons2'
      le_list_drop_Cons_neq ~> Sublist.sublisteq_Cons2_neq
      le_list_drop_Cons ~> Sublist.sublisteq_Cons'
      le_list_drop_many ~> Sublist.sublisteq_drop_many
      le_list_filter_left ~> Sublist.sublisteq_filter_left
      le_list_rev_drop_many ~> Sublist.sublisteq_rev_drop_many
      le_list_rev_take_iff ~> Sublist.sublisteq_append
      le_list_same_length ~> Sublist.sublisteq_same_length
      le_list_take_many_iff ~> Sublist.sublisteq_append'
      less_eq_list.drop ~> less_eq_list_drop
      less_eq_list.induct ~> less_eq_list_induct
      not_le_list_length ~> Sublist.not_sublisteq_length

INCOMPATIBILITY.

* New theory Library/Countable_Set.

* Theory Library/Debug and Library/Parallel provide debugging and
parallel execution for code generated towards Isabelle/ML.

* Theory Library/FuncSet: Extended support for Pi and extensional and
introduce the extensional dependent function space "PiE". Replaced
extensional_funcset by an abbreviation, and renamed lemmas from
extensional_funcset to PiE as follows:

  extensional_empty  ~>  PiE_empty
  extensional_funcset_empty_domain  ~>  PiE_empty_domain
  extensional_funcset_empty_range  ~>  PiE_empty_range
  extensional_funcset_arb  ~>  PiE_arb
  extensional_funcset_mem  ~>  PiE_mem
  extensional_funcset_extend_domainI  ~>  PiE_fun_upd
  extensional_funcset_restrict_domain  ~>  fun_upd_in_PiE
  extensional_funcset_extend_domain_eq  ~>  PiE_insert_eq
  card_extensional_funcset  ~>  card_PiE
  finite_extensional_funcset  ~>  finite_PiE

INCOMPATIBILITY.

* Theory Library/FinFun: theory of almost everywhere constant
functions (supersedes the AFP entry "Code Generation for Functions as
Data").

* Theory Library/Phantom: generic phantom type to make a type
parameter appear in a constant's type.  This alternative to adding
TYPE('a) as another parameter avoids unnecessary closures in generated
code.

* Theory Library/RBT_Impl: efficient construction of red-black trees
from sorted associative lists. Merging two trees with rbt_union may
return a structurally different tree than before.  Potential
INCOMPATIBILITY.

* Theory Library/IArray: immutable arrays with code generation.

* Theory Library/Finite_Lattice: theory of finite lattices.

* HOL/Multivariate_Analysis: replaced

  "basis :: 'a::euclidean_space => nat => real"
  "\<Chi>\<Chi> :: (nat => real) => 'a::euclidean_space"

on euclidean spaces by using the inner product "_ \<bullet> _" with
vectors from the Basis set: "\<Chi>\<Chi> i. f i" is superseded by
"SUM i : Basis. f i * r i".

  With this change the following constants are also changed or removed:

    DIM('a) :: nat  ~>  card (Basis :: 'a set)   (is an abbreviation)
    a $$ i  ~>  inner a i  (where i : Basis)
    cart_base i  removed
    \<pi>, \<pi>'  removed

  Theorems about these constants where removed.

  Renamed lemmas:

    component_le_norm  ~>  Basis_le_norm
    euclidean_eq  ~>  euclidean_eq_iff
    differential_zero_maxmin_component  ~>  differential_zero_maxmin_cart
    euclidean_simps  ~>  inner_simps
    independent_basis  ~>  independent_Basis
    span_basis  ~>  span_Basis
    in_span_basis  ~>  in_span_Basis
    norm_bound_component_le  ~>  norm_boound_Basis_le
    norm_bound_component_lt  ~>  norm_boound_Basis_lt
    component_le_infnorm  ~>  Basis_le_infnorm

INCOMPATIBILITY.

* HOL/Probability:

  - Added simproc "measurable" to automatically prove measurability.

  - Added induction rules for sigma sets with disjoint union
    (sigma_sets_induct_disjoint) and for Borel-measurable functions
    (borel_measurable_induct).

  - Added the Daniell-Kolmogorov theorem (the existence the limit of a
    projective family).

* HOL/Cardinals: Theories of ordinals and cardinals (supersedes the
AFP entry "Ordinals_and_Cardinals").

* HOL/BNF: New (co)datatype package based on bounded natural functors
with support for mixed, nested recursion and interesting non-free
datatypes.

* HOL/Finite_Set and Relation: added new set and relation operations
expressed by Finite_Set.fold.

* New theory HOL/Library/RBT_Set: implementation of sets by red-black
trees for the code generator.

* HOL/Library/RBT and HOL/Library/Mapping have been converted to
Lifting/Transfer.
possible INCOMPATIBILITY.

* HOL/Set: renamed Set.project -> Set.filter
INCOMPATIBILITY.


*** Document preparation ***

* Dropped legacy antiquotations "term_style" and "thm_style", since
styles may be given as arguments to "term" and "thm" already.
Discontinued legacy styles "prem1" .. "prem19".

* Default LaTeX rendering for \<euro> is now based on eurosym package,
instead of slightly exotic babel/greek.

* Document variant NAME may use different LaTeX entry point
document/root_NAME.tex if that file exists, instead of the common
document/root.tex.

* Simplified custom document/build script, instead of old-style
document/IsaMakefile.  Minor INCOMPATIBILITY.


*** ML ***

* The default limit for maximum number of worker threads is now 8,
instead of 4, in correspondence to capabilities of contemporary
hardware and Poly/ML runtime system.

* Type Seq.results and related operations support embedded error
messages within lazy enumerations, and thus allow to provide
informative errors in the absence of any usable results.

* Renamed Position.str_of to Position.here to emphasize that this is a
formal device to inline positions into message text, but not
necessarily printing visible text.


*** System ***

* Advanced support for Isabelle sessions and build management, see
"system" manual for the chapter of that name, especially the "isabelle
build" tool and its examples.  The "isabelle mkroot" tool prepares
session root directories for use with "isabelle build", similar to
former "isabelle mkdir" for "isabelle usedir".  Note that this affects
document preparation as well.  INCOMPATIBILITY, isabelle usedir /
mkdir / make are rendered obsolete.

* Discontinued obsolete Isabelle/build script, it is superseded by the
regular isabelle build tool.  For example:

  isabelle build -s -b HOL

* Discontinued obsolete "isabelle makeall".

* Discontinued obsolete IsaMakefile and ROOT.ML files from the
Isabelle distribution, except for rudimentary src/HOL/IsaMakefile that
provides some traditional targets that invoke "isabelle build".  Note
that this is inefficient!  Applications of Isabelle/HOL involving
"isabelle make" should be upgraded to use "isabelle build" directly.

* The "isabelle options" tool prints Isabelle system options, as
required for "isabelle build", for example.

* The "isabelle logo" tool produces EPS and PDF format simultaneously.
Minor INCOMPATIBILITY in command-line options.

* The "isabelle install" tool has now a simpler command-line.  Minor
INCOMPATIBILITY.

* The "isabelle components" tool helps to resolve add-on components
that are not bundled, or referenced from a bare-bones repository
version of Isabelle.

* Settings variable ISABELLE_PLATFORM_FAMILY refers to the general
platform family: "linux", "macos", "windows".

* The ML system is configured as regular component, and no longer
picked up from some surrounding directory.  Potential INCOMPATIBILITY
for home-made settings.

* Improved ML runtime statistics (heap, threads, future tasks etc.).

* Discontinued support for Poly/ML 5.2.1, which was the last version
without exception positions and advanced ML compiler/toplevel
configuration.

* Discontinued special treatment of Proof General -- no longer guess
PROOFGENERAL_HOME based on accidental file-system layout.  Minor
INCOMPATIBILITY: provide PROOFGENERAL_HOME and PROOFGENERAL_OPTIONS
settings manually, or use a Proof General version that has been
bundled as Isabelle component.



New in Isabelle2012 (May 2012)
------------------------------

*** General ***

* Prover IDE (PIDE) improvements:

  - more robust Sledgehammer integration (as before the sledgehammer
    command-line needs to be typed into the source buffer)
  - markup for bound variables
  - markup for types of term variables (displayed as tooltips)
  - support for user-defined Isar commands within the running session
  - improved support for Unicode outside original 16bit range
    e.g. glyph for \<A> (thanks to jEdit 4.5.1)

* Forward declaration of outer syntax keywords within the theory
header -- minor INCOMPATIBILITY for user-defined commands.  Allow new
commands to be used in the same theory where defined.

* Auxiliary contexts indicate block structure for specifications with
additional parameters and assumptions.  Such unnamed contexts may be
nested within other targets, like 'theory', 'locale', 'class',
'instantiation' etc.  Results from the local context are generalized
accordingly and applied to the enclosing target context.  Example:

  context
    fixes x y z :: 'a
    assumes xy: "x = y" and yz: "y = z"
  begin

  lemma my_trans: "x = z" using xy yz by simp

  end

  thm my_trans

The most basic application is to factor-out context elements of
several fixes/assumes/shows theorem statements, e.g. see
~~/src/HOL/Isar_Examples/Group_Context.thy

Any other local theory specification element works within the "context
... begin ... end" block as well.

* Bundled declarations associate attributed fact expressions with a
given name in the context.  These may be later included in other
contexts.  This allows to manage context extensions casually, without
the logical dependencies of locales and locale interpretation.  See
commands 'bundle', 'include', 'including' etc. in the isar-ref manual.

* Commands 'lemmas' and 'theorems' allow local variables using 'for'
declaration, and results are standardized before being stored.  Thus
old-style "standard" after instantiation or composition of facts
becomes obsolete.  Minor INCOMPATIBILITY, due to potential change of
indices of schematic variables.

* Rule attributes in local theory declarations (e.g. locale or class)
are now statically evaluated: the resulting theorem is stored instead
of the original expression.  INCOMPATIBILITY in rare situations, where
the historic accident of dynamic re-evaluation in interpretations
etc. was exploited.

* New tutorial "Programming and Proving in Isabelle/HOL"
("prog-prove").  It completely supersedes "A Tutorial Introduction to
Structured Isar Proofs" ("isar-overview"), which has been removed.  It
also supersedes "Isabelle/HOL, A Proof Assistant for Higher-Order
Logic" as the recommended beginners tutorial, but does not cover all
of the material of that old tutorial.

* Updated and extended reference manuals: "isar-ref",
"implementation", "system"; reduced remaining material in old "ref"
manual.


*** Pure ***

* Command 'definition' no longer exports the foundational "raw_def"
into the user context.  Minor INCOMPATIBILITY, may use the regular
"def" result with attribute "abs_def" to imitate the old version.

* Attribute "abs_def" turns an equation of the form "f x y == t" into
"f == %x y. t", which ensures that "simp" or "unfold" steps always
expand it.  This also works for object-logic equality.  (Formerly
undocumented feature.)

* Sort constraints are now propagated in simultaneous statements, just
like type constraints.  INCOMPATIBILITY in rare situations, where
distinct sorts used to be assigned accidentally.  For example:

  lemma "P (x::'a::foo)" and "Q (y::'a::bar)"  -- "now illegal"

  lemma "P (x::'a)" and "Q (y::'a::bar)"
    -- "now uniform 'a::bar instead of default sort for first occurrence (!)"

* Rule composition via attribute "OF" (or ML functions OF/MRS) is more
tolerant against multiple unifiers, as long as the final result is
unique.  (As before, rules are composed in canonical right-to-left
order to accommodate newly introduced premises.)

* Renamed some inner syntax categories:

    num ~> num_token
    xnum ~> xnum_token
    xstr ~> str_token

Minor INCOMPATIBILITY.  Note that in practice "num_const" or
"num_position" etc. are mainly used instead (which also include
position information via constraints).

* Simplified configuration options for syntax ambiguity: see
"syntax_ambiguity_warning" and "syntax_ambiguity_limit" in isar-ref
manual.  Minor INCOMPATIBILITY.

* Discontinued configuration option "syntax_positions": atomic terms
in parse trees are always annotated by position constraints.

* Old code generator for SML and its commands 'code_module',
'code_library', 'consts_code', 'types_code' have been discontinued.
Use commands of the generic code generator instead.  INCOMPATIBILITY.

* Redundant attribute "code_inline" has been discontinued. Use
"code_unfold" instead.  INCOMPATIBILITY.

* Dropped attribute "code_unfold_post" in favor of the its dual
"code_abbrev", which yields a common pattern in definitions like

  definition [code_abbrev]: "f = t"

INCOMPATIBILITY.

* Obsolete 'types' command has been discontinued.  Use 'type_synonym'
instead.  INCOMPATIBILITY.

* Discontinued old "prems" fact, which used to refer to the accidental
collection of foundational premises in the context (already marked as
legacy since Isabelle2011).


*** HOL ***

* Type 'a set is now a proper type constructor (just as before
Isabelle2008).  Definitions mem_def and Collect_def have disappeared.
Non-trivial INCOMPATIBILITY.  For developments keeping predicates and
sets separate, it is often sufficient to rephrase some set S that has
been accidentally used as predicates by "%x. x : S", and some
predicate P that has been accidentally used as set by "{x. P x}".
Corresponding proofs in a first step should be pruned from any
tinkering with former theorems mem_def and Collect_def as far as
possible.

For developments which deliberately mix predicates and sets, a
planning step is necessary to determine what should become a predicate
and what a set.  It can be helpful to carry out that step in
Isabelle2011-1 before jumping right into the current release.

* Code generation by default implements sets as container type rather
than predicates.  INCOMPATIBILITY.

* New type synonym 'a rel = ('a * 'a) set

* The representation of numerals has changed.  Datatype "num"
represents strictly positive binary numerals, along with functions
"numeral :: num => 'a" and "neg_numeral :: num => 'a" to represent
positive and negated numeric literals, respectively.  See also
definitions in ~~/src/HOL/Num.thy.  Potential INCOMPATIBILITY, some
user theories may require adaptations as follows:

  - Theorems with number_ring or number_semiring constraints: These
    classes are gone; use comm_ring_1 or comm_semiring_1 instead.

  - Theories defining numeric types: Remove number, number_semiring,
    and number_ring instances. Defer all theorems about numerals until
    after classes one and semigroup_add have been instantiated.

  - Numeral-only simp rules: Replace each rule having a "number_of v"
    pattern with two copies, one for numeral and one for neg_numeral.

  - Theorems about subclasses of semiring_1 or ring_1: These classes
    automatically support numerals now, so more simp rules and
    simprocs may now apply within the proof.

  - Definitions and theorems using old constructors Pls/Min/Bit0/Bit1:
    Redefine using other integer operations.

* Transfer: New package intended to generalize the existing
"descending" method and related theorem attributes from the Quotient
package.  (Not all functionality is implemented yet, but future
development will focus on Transfer as an eventual replacement for the
corresponding parts of the Quotient package.)

  - transfer_rule attribute: Maintains a collection of transfer rules,
    which relate constants at two different types. Transfer rules may
    relate different type instances of the same polymorphic constant,
    or they may relate an operation on a raw type to a corresponding
    operation on an abstract type (quotient or subtype). For example:

    ((A ===> B) ===> list_all2 A ===> list_all2 B) map map
    (cr_int ===> cr_int ===> cr_int) (%(x,y) (u,v). (x+u, y+v)) plus_int

  - transfer method: Replaces a subgoal on abstract types with an
    equivalent subgoal on the corresponding raw types. Constants are
    replaced with corresponding ones according to the transfer rules.
    Goals are generalized over all free variables by default; this is
    necessary for variables whose types change, but can be overridden
    for specific variables with e.g. "transfer fixing: x y z".  The
    variant transfer' method allows replacing a subgoal with one that
    is logically stronger (rather than equivalent).

  - relator_eq attribute: Collects identity laws for relators of
    various type constructors, e.g. "list_all2 (op =) = (op =)".  The
    transfer method uses these lemmas to infer transfer rules for
    non-polymorphic constants on the fly.

  - transfer_prover method: Assists with proving a transfer rule for a
    new constant, provided the constant is defined in terms of other
    constants that already have transfer rules. It should be applied
    after unfolding the constant definitions.

  - HOL/ex/Transfer_Int_Nat.thy: Example theory demonstrating transfer
    from type nat to type int.

* Lifting: New package intended to generalize the quotient_definition
facility of the Quotient package; designed to work with Transfer.

  - lift_definition command: Defines operations on an abstract type in
    terms of a corresponding operation on a representation
    type.  Example syntax:

    lift_definition dlist_insert :: "'a => 'a dlist => 'a dlist"
      is List.insert

    Users must discharge a respectfulness proof obligation when each
    constant is defined. (For a type copy, i.e. a typedef with UNIV,
    the proof is discharged automatically.) The obligation is
    presented in a user-friendly, readable form; a respectfulness
    theorem in the standard format and a transfer rule are generated
    by the package.

  - Integration with code_abstype: For typedefs (e.g. subtypes
    corresponding to a datatype invariant, such as dlist),
    lift_definition generates a code certificate theorem and sets up
    code generation for each constant.

  - setup_lifting command: Sets up the Lifting package to work with a
    user-defined type. The user must provide either a quotient theorem
    or a type_definition theorem.  The package configures transfer
    rules for equality and quantifiers on the type, and sets up the
    lift_definition command to work with the type.

  - Usage examples: See Quotient_Examples/Lift_DList.thy,
    Quotient_Examples/Lift_RBT.thy, Quotient_Examples/Lift_FSet.thy,
    Word/Word.thy and Library/Float.thy.

* Quotient package:

  - The 'quotient_type' command now supports a 'morphisms' option with
    rep and abs functions, similar to typedef.

  - 'quotient_type' sets up new types to work with the Lifting and
    Transfer packages, as with 'setup_lifting'.

  - The 'quotient_definition' command now requires the user to prove a
    respectfulness property at the point where the constant is
    defined, similar to lift_definition; INCOMPATIBILITY.

  - Renamed predicate 'Quotient' to 'Quotient3', and renamed theorems
    accordingly, INCOMPATIBILITY.

* New diagnostic command 'find_unused_assms' to find potentially
superfluous assumptions in theorems using Quickcheck.

* Quickcheck:

  - Quickcheck returns variable assignments as counterexamples, which
    allows to reveal the underspecification of functions under test.
    For example, refuting "hd xs = x", it presents the variable
    assignment xs = [] and x = a1 as a counterexample, assuming that
    any property is false whenever "hd []" occurs in it.

    These counterexample are marked as potentially spurious, as
    Quickcheck also returns "xs = []" as a counterexample to the
    obvious theorem "hd xs = hd xs".

    After finding a potentially spurious counterexample, Quickcheck
    continues searching for genuine ones.

    By default, Quickcheck shows potentially spurious and genuine