ViewVC logotype

Log of /metaprl/theories/tactic/top_conversionals.ml

Parent Directory Parent Directory | Revision Log Revision Log

Sticky Revision:
(Current path doesn't exist after revision 4538)

Revision 3584 - (view) (download) (annotate) - [select for diffs]
Modified Thu Apr 25 15:28:40 2002 UTC (19 years, 2 months ago) by nogin
File length: 15958 byte(s)
Diff to previous 3329
- Added the comment module to the theories.pdf ("make latex").

- Added a hack allowing to override a theory display with another theory's
display. This is necessary for bootstrapping reasons - it allows one
to get a reasonable output of a theory (e.g. Comment) that is defined before
all essential display forms (e.g. the Summary ones) are there.

- Added debugging code to make it easier to find dforms that cause
zone begin/end mismatch (and used it to get rid of all
the "Unballanced buffer" warnings in "make latex").

- Minor dforms&comments updates.

Revision 3329 - (view) (download) (annotate) - [select for diffs]
Modified Fri Jul 13 21:34:19 2001 UTC (19 years, 11 months ago) by nogin
File length: 15962 byte(s)
Diff to previous 3297
I've removed old Itt_int* theories and switched all the ITT to Yegor's
new theories.

Revision 3297 - (view) (download) (annotate) - [select for diffs]
Modified Sun Jun 24 21:45:02 2001 UTC (20 years ago) by nogin
File length: 15961 byte(s)
Diff to previous 3292
Resource intermediate data type is irrelevant, so they are no longer required
in the resource declaration.

Revision 3292 - (view) (download) (annotate) - [select for diffs]
Modified Sun Jun 24 10:25:49 2001 UTC (20 years ago) by nogin
File length: 16000 byte(s)
Diff to previous 3283
*** IMPORTANT: before doing "cvs update" see warning below ***

This commit implements a global resource that brings together resource management
that used to be done by each resource in its own way.

- A much simplier object (Mp_resource.resource_info) is now needed to create a new resource.
Additionally, many resource helper modules, such as term_match_table now provide functions
that create a resource_info for you.

- Funtions get_tactic_arg, get_int_tactic_arg, etc were replaced by a single function
get_resource_arg (note: functions for basci types, such as get_with_arg, get_sel_args, etc
were _not_ affected)

- Resource declaration statement now declares and resource creation statement now creates
a function get_<name>_resource. This function ca be passed as a second argument
to get_resource_arg to retrieve the current value of a particular resource.

- Resource annotation functions are now separate from the resource declaration/creation
mechanism. Resource annotation on a rule are now passed to a function named
process_<name>_resource_annotation. This function must have the appropriate
Mp_resource.annotation_processor type for the resource. This function can be declared
and implemented in _any_ module, not just the module that declares and creates the
<name> resource.

- Currenlty MetaPRL leaves it to Ocaml to keep track of these functions.
Consequently the module that defines the process_<name>_resource_annotation function
must be open'ed (not included!) by modules that use <name> annotations.


Unfinished business in this commit:
- Remainder of TODO 1.12 and 1.13

- Some proofs were broken by this commit. It appears that they were broken for
a "good" reason, but still need to be fixed.

- I need to make sure things now always have the right precedences (all things being equal,
the later an item is added to resource, the earlier precedence it should have).

- Consider adding refiner sentinels and display forms to global resource as well.
Currently each of them still does its own way of resource management (although the way
display forms are managed was changed by this commit - need to make sure the precedences
are right).

- Distributed refinement: I am afraid that the current resource code
will either work correctly only if all processes in the distributed group have
the same global resource, or it will attempt to pass around all the resource data.
Obviously, neither of these choices is a good one.

*** WARNING ***

This commit changes the tactic_arg type and also makes minor changes in
the FilterCache.info type. Because of that old .prlb and .cmoz files
*will no longer work*. Old .prla files will work, but will produce lots
of (harmless) Filter_summary.dest_term warning.

Before doing "cvs update" do the following:
1) export all your unsaved proofs into .prla files.
2) rm -f theories/*/*.prlb
3) run "make clean"

Revision 3283 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 20 19:21:25 2001 UTC (20 years ago) by kopylov
File length: 16892 byte(s)
Diff to previous 3282
Fix an error

Revision 3282 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 20 18:41:55 2001 UTC (20 years ago) by kopylov
File length: 16822 byte(s)
Diff to previous 3279
New tactics ans cnversionals are documented.

Revision 3279 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 20 11:48:09 2001 UTC (20 years ago) by nogin
File length: 15346 byte(s)
Diff to previous 3278
- Renamed andthenC -> thenC
- Added byDefT : conv -> tactic
- Proved rewrites in Itt_dprod
- Removed some unused stuff from one of the filter_prog interfaces.

Revision 3278 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 20 00:27:36 2001 UTC (20 years ago) by nogin
File length: 15361 byte(s)
Diff to previous 3277
- Changed the syntax of resource implementation. Now you no longer need to
re-declare resource types in the implementation. Instead of the declaration
and a Mp_resource.create call, use the following syntax in the implementation:

let resource <name> = <expr>

where <expr> is an expression that of the type Mp_resource.info

- Additionally, it is no longer necessary to use _resource in resource names
and resource annotations.

Revision 3277 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 20 00:12:25 2001 UTC (20 years ago) by kopylov
File length: 15469 byte(s)
Diff to previous 3276
- New tacticals
    onAllAssumT : (int -> tactic) -> tactic
    onAllMAssumT : (int -> tactic) -> tactic
  apply a tactic to all assumtions

 rwAll applies a convertion to all clauses
 rwcAll applies a convertion to all clauses for the given assumption
 rwAllAll applies a convertion to all assumption and to the goal sequent

 rwhAll,rwchAll,rwhAllAll, rwaAll,rwcaAll,rwaAllAll work by analogy

 Now these tactics does not work properly (see BUGS 3.12-3.14)

- Most of the theorems on records are proved

Revision 3276 - (view) (download) (annotate) - [select for diffs]
Modified Tue Jun 19 15:17:10 2001 UTC (20 years ago) by nogin
File length: 14992 byte(s)
Diff to previous 3257
Added support for the following syntax of resource improvement:
let resource name += expr
let resource name += [ expr1; expr2; ...; exprn ]

Revision 3257 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jun 13 19:32:30 2001 UTC (20 years ago) by kopylov
File length: 15131 byte(s)
Diff to previous 3235
1. I added new primitive rules:

   The last two rules show the differences between pointwise and pairwise functionality.
   They are valid in pointwise functionality (as well as quoationtElimination2).

2. Itt_struct3 contains some rules that derived from these new ones.
   But thesd rules are valid in both functionalities.

3. I moved definition of natural numbers to a new theory Itt_nat.
   To prove induction for natural numbers one needs Itt_struct3.
   I almost proved it (leaving some simple arithmetic subgoals, that should
   be killed by arith)

4. I defined record type.
   Itt_record is a main theory of records.
   Itt_record_exm contains some examples.

5. I add a new conversion
        * applyAllC : conv list -> conv
   that applies all conversions to all subterms and a tactic
        * rwa : conv list -> int -> tactic
        = rw (applyAllC convs)

Revision 3235 - (view) (download) (annotate) - [select for diffs]
Modified Sun May 20 23:04:08 2001 UTC (20 years, 1 month ago) by nogin
File length: 15039 byte(s)
Diff to previous 3212
Display form fixes.

- TeX: all object-level terms (theorem statements, etc) are now typeset
in math mode.

- Tex: variable names are now parsed to detect indeces. In particular,
  'a1 will now be printed as a_{1} and 'a1_2 will be printed as a_{1,2}.

- Tex: `|' symbol in display forms is now always handled correctly.

- Numerous small fixes.

Revision 3212 - (view) (download) (annotate) - [select for diffs]
Modified Sun May 6 21:08:07 2001 UTC (20 years, 1 month ago) by nogin
File length: 14989 byte(s)
Diff to previous 3193
A few display form fixes.

Revision 3193 - (view) (download) (annotate) - [select for diffs]
Modified Sat Mar 24 18:03:11 2001 UTC (20 years, 3 months ago) by nogin
File length: 14977 byte(s)
Diff to previous 3125
Fixed the show_loading statements to ensure we do not have the same ones
in different files.
Thanks to Morten <mrhiger@daimi.au.dk> for noticing this problem.

Revision 3125 - (view) (download) (annotate) - [select for diffs]
Modified Thu Feb 1 01:18:20 2001 UTC (20 years, 4 months ago) by kopylov
File length: 14969 byte(s)
Diff to previous 3124
1. The untilFailT tactical is implemented.

2. Now there are the untilFailC, whileProgressC and repeatC conversionals
hat work as corresponding tacticals.

3. The failC a failWithC conversions now work as failT and failWithT

4. repeatForT and repeatForC now raise an exception when applied to
a negative argument.

Revision 3124 - (view) (download) (annotate) - [select for diffs]
Modified Wed Jan 31 22:44:31 2001 UTC (20 years, 4 months ago) by nogin
File length: 14407 byte(s)
Diff to previous 3058
Spelling changes.
For best results, upgrade to http://www.cs.cornell.edu/nogin/RPM/MetaPRL/noarch/words-2.1-1.noarch.html

Revision 3058 - (view) (download) (annotate) - [select for diffs]
Modified Sun Sep 10 20:41:19 2000 UTC (20 years, 9 months ago) by jyh
File length: 14403 byte(s)
Diff to previous 2822
Last commit failed partway through...

Revision 2822 - (view) (download) (annotate) - [select for diffs]
Modified Fri Oct 22 01:07:59 1999 UTC (21 years, 8 months ago) by nogin
File length: 5719 byte(s)
Diff to previous 2799
Added new function to mp_debug
let show_loading s = if !debug_load then Printf.eprintf s eflush

and replaced all usages of debug_load with show_loading

Revision 2799 - (view) (download) (annotate) - [select for diffs]
Modified Fri Jul 23 00:24:57 1999 UTC (21 years, 11 months ago) by nogin
File length: 5747 byte(s)
Diff to previous 2743
Now one can use SO variables in display form specifications

Revision 2743 - (view) (download) (annotate) - [select for diffs]
Added Wed Jun 23 04:50:29 1999 UTC (22 years ago) by jyh
File length: 5781 byte(s)
This is a major release; I bumped the release number to 0.6.
WARNING: There are major changes here, and there may be problems
   with this version that prevent you from using your old proof.
   Save your work before using this version.
NOTE: If you decide to try this out, and you find problems,
   please(!) send me email so I can fix it.

   1. The proof structure is totally changed, so that the proof
      editor edits the raw proof extracts.  This means that you
      can view the actions of the refiner down to the primitive
      rule applications.  In the proof editor, you can use
      "down 0" (or "cd "0") to descend into the proof of a rule box.
      Primitive proofs are called "extracts", and are labeled with
      brackets (like [extract]).  To expand the extract, use the command
      "unfold ()".  You should be able to get all the way down to
      the primitive rule/rewrite applications.

      This also means that the format of the proof files (the .prlb
      files) has changed.  The old proof files are still valid,
      but this is a hack and will be deprecated in the next
      few months.  I haven't yet added Alexey's ASCII proof
      files, but that will come with the next release.

      As usual, the "undo ()" command undoes the last proof step,
      including "unfold ()".  The "nop ()" command, reset the undo
      stack.  I also added a "redo ()" command that undoes the
      last undo.

      There is a simple rule-box proof cache for collecting proofs
      as you edit them.  If cached proofs are available, you will
      see them in brackets (like [* # * *]) on the status line.
      I haven't yet:( added the commands to use cached proofs.

   2. Refiner changes.  I added two new features: the "cutT <term>"
      tactic cuts in a new assumption.  Also, you can apply
      rewrites directly on assumptions, with the "rwc" and
      "rwch" operations, that take a clause argument.  Basically,
      this means that instead of folding the goal, you can unfold
      the assumption.  I believe this is sound; let me know if
      you think otherwise!

   3. Resource changes.  I added resource automation, built on
      the basic resource parsing Alexey added.  Ratherthan writing
      resource code for every rule, you can annotate most rules
      to generate the resource code directly.  You can see lots of
      examples in the Itt theory.  Basically, there are three useful
      resources annotations:
         intro_resource []: adds the rule as an introduction in dT
         intro_resource [SelectOption i]: adds to selT i dT
         elim_resource []: adds as an elimination rule to dT.
            This normally tries to thin the hyp that was eliminated.
         elim_resource [ThinOption]: don't thin the hyp

      Rules should be annotated with labels on their clauses,
      like [wf], [main], [assertion], etc.  This means that in
      most tactic aplcations, you no longer need to have the
      thenLT [addHiddenLabel "main"; ...] part.

      N.B.  This is the most likey parts of this release to
      cause problems, because I deleted most old resource code.

      Also, you can still write standard resource code, but there
      is no longer a d_resource--it has been broken into two parts:
      the intro_resource for application to goals, and elim_resource
      for application to hyps.

   4. The proof editor is multi-threaded, so you can work on multiple
      proofs simultaneously.  In the normal Emacs editor, this is
      no help for you.  But there is a new Java editor with the
      standard point-and-click interface, and it views the proof
      in HTML, with multiple windows etc.  The beautiful thing is
      that you can use display forms to add commands to edit windows.
      The sad thing is that I built it on NT, Java 1.2 is required,
      and I haven't tried the blackdown.org Java release on Linux.
      This editor is pending some bug fixes from Sun to get the
      fonts right (they call this a standard release?).

      In any case, here are the direct implications.  The display
      forms have an "html" mode.  The display form formatting in
      the module Rformat has been completely redone, but display
      _should_ be the same as it was before.

      It is likely that I will add an HTML dump, so we can send
      uneditable proofs around by email or on the web.  Check out
      the file theories/base/summary.ml to see some example HTML
      display forms.

      The other issue: your MetaPRL process starts a web server on
      YOUR local machine using YOUR process id on the "next" available
      TCP port, and it serves files only from the search path that you pass
      MetaPRL.  I realize that this has security implications.  This
      is necessary to get browser access to the working MetaPRL proof.
      Is this crazy?  Let me know your beliefs, religious or

   5. There are numerous minor changes.  I redid parts of the WeakMemo,
      Term_header_constr, and TermHash.  The theories/tactic directory
      has been split into tactic/boot (which is not compiled by MetaPRL),
      and theories/tactic (which is).


This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, enter a numeric revision.

  Diffs between and
  Type of Diff should be a

  ViewVC Help
Powered by ViewVC 1.1.26