tuning

finished splitting sum types for corecursors

split sum types in corecursor definition

first step towards splitting corecursor function arguments into (p, g, h) triples

reverted "id" change: The problem is rather that the "%c. f c" argument sometimes gets eta-reduced

replaced jedit_relative_font_size by jedit_font_scale;

need to provide label via some jEdit property;

some support to organize options in sections;

merged

generate "id" rather than (%v. v)

correctly generate sel_coiter and sel_corec theorems

generate "sel_coiters" and friends

sanity check

implemented and use "mk_sum_casesN_balanced"

fixed general case of "mk_sumEN_balanced"

debug

fixed base case of "mk_sumEN_balanced"

prevent inconsistent selector types

minor optimization

allow same selector name for several constructors

removed done TODO

avoid type inference + tuning

use balanced sums for constructors (to gracefully handle 100 constructors or more)

busted -- let's use more neutral names

replacing own dummy value by Haskell's Prelude.undefined

prefer global default font over IsabelleText of jEdit TextArea;

uniform operation on initial delay;

option jedit_load_delay; tuned;

dynamic evaluation of time (e.g. via options); proper setInitialDelay, not setDelay;

proper multi-line tooltip;

more detailed option tooltip; more formal option.load; properties change propagation to Session_Dockable;

more systematic JEdit_Options.make_component; separate module Isabelle_Logic;

manage Isabelle/jEdit options as Isabelle/Scala options (with persistent preferences);

formal markup for @{file} (for hyperlinks etc.) -- interpret path wrt. master directory as usual;

more explicit indication of legacy features;

more explicit indication of legacy features;

simplify "Process" example even further

stabilized generation of parameterized theorem

added snippets

simplify "Process" example further

simplify "Process" example

full name of a type as key in bnf table

fixed bug with one-value codatatype "codata 'a dead_foo = A"

tuning

fixed and reenabled "corecs" theorems

fixed and enabled generation of "coiters" theorems, including the recursive case

generate "fld_unf_corecs" as well

reactivated generation of "coiters" theorems

use map_id, not map_id', to allow better composition

open typedefs everywhere in the package

open typedef for datatypes

fixed and enabled iterator/recursor theorems

renamed for consistency

oops

tuning

for compatiblity with old datatype package: not only "recs" with "s", but also "iters" and their "fld_"/"unf_" variants

fixed bug with one-value types with phantom type arguments

imported patch debugging

repaired "nofail4" example

renamed xxxBNF to pre_xxx

fixed handling of map of "fun"

comment out code that's not ready

tuning

construct the right iterator theorem in the recursive case

some work on coiter tactic

more sugar on codatatypes

define corecursors

define coiterators

TODO

tuning

completed iter/rec proofs

TODOs

implemented "mk_iter_or_rec_tac"

generate iter/rec goals

repaired constant types

some work towards iterator and recursor properties

tuning

correctly curry recursor arguments

added high-level recursor, not yet curried

merged

tuned proofs;

postpone update of text overview panel after incoming session edits, to improve reactivity of editing massive theories like src/HOL/Multivariate_Analysis;

more explicit Delay operations;

tuned proofs;

clearer names for functions in Quickcheck's narrowing engine

merged

tuned latex

lattice instances for option type

combinator Option.these

adjusted examples

countable_datatype method: pre-instantiate induction rule to avoid failure with e.g. datatype a = A "b list" and b = B "a"

handle type constructors not known to be a BNF using the DEADID BNF

respect order of/additional type variables supplied by the user in fixed point constructions;

gracefully handle shadowing case (fourth step of sugar localization)

careful about constructor types w.r.t. fake context (third step of localization)

read the real types off the constant types, rather than using the fake parser types (second step of sugar localization)

tuning

use "add_type" rather than "add_types_global"

don't throw away the context when hacking the theory (first step to localize the sugar code)

tuning

introduced and used "mk_Freesss", and simplified "mk_Freess(')"

construct high-level iterator RHS

option for discarding build results, enabled in particular for Isabelle_makeall

by default, only generate one discriminator for a two-value datatype

eliminated potentially confusing terminology of Scala "layer";

merged

tuned proofs;

discontinued obsolete fork_mono to loosen some brakes -- NB: TTY interaction has Goal.future_proofs disabled due to missing Future.worker_task;

honor mixfix specifications

code indentation

print timing information

adapted example