		Search-engine friendly clone of the ACL2 documentation.

Defttag

Set-raw-mode

Enter or exit ``raw mode,'' a raw Lisp environment

Below we discuss raw-mode. In brief: The simplest way to turn raw-mode on is :SET-RAW-MODE-ON!, and to turn it off, :SET-RAW-MODE NIL. Also see set-raw-mode-on!.

ACL2 users often find its careful syntax checking to be helpful during code development. Sometimes it is even useful to do code development in :logic mode, where ACL2 can be used to check termination of (mutually) recursive functions, verify guards, or even prove properties of the functions.

However, loading code using include-book is much slower than using Common Lisp load in raw Lisp, and in this sense ACL2 can get in the way of efficient execution. Unfortunately, it is error-prone to use ACL2 sources (or their compilations) in raw Lisp, primarily because a number of ACL2 primitives will not let you do so. Perhaps you have seen this error message when trying to do so:

HARD ACL2 ERROR in ACL2-UNWIND-PROTECT:  Apparently you have tried
to execute a form in raw Lisp that is only intended to be executed
inside the ACL2 loop.

Even without this problem it is important to enter the ACL2 loop (see lp), for example in order to set the cbd and (to get more technical) the readtable.

ACL2 provides a ``raw mode'' for execution of raw Lisp forms. In this mode, include-book reduces essentially to a Common Lisp load. More generally, the ACL2 logical world is not routinely extended in raw mode (some sneaky tricks are probably required to make that happen). To turn raw mode off or on:

:set-raw-mode t   ; turn raw mode on
:set-raw-mode-on! ; same as above, but no trust tag required
:set-raw-mode nil ; turn raw mode off

The way you can tell that you are in raw mode is by looking at the prompt (see default-print-prompt), which uses a capital ``P'' (suggesting something like program mode, but even further from logic mode).

ACL2 P>

Typical benefits of raw mode are fast loading of source and compiled files and the capability to hack arbitrary Common Lisp code in an environment with the ACL2 sources loaded (and hence with ACL2 primitives available). However, it probably is generally best to avoid raw mode unless these advantages seem important. We expect the main benefit of raw mode to be in deployment of applications, where raw Lisp code may be useful, and where load time is much faster than the time required for a full-blown include-book — but not that the fast loading of books and treatment of hard errors discussed above may be useful during development.

Raw mode is also useful for those who want to build extensions of ACL2. For example, the following form can be put into a certifiable book to load an arbitrary Common Lisp source or compiled file.

(progn (defttag my-application)
       (progn! (set-raw-mode t)
               (load "some-file")))

Also see include-raw with-raw-mode, defttag, and progn!.

Below are several disadvantages to using raw mode. These should discourage users from using it for general code development, as :program mode is generally preferable.

Forms are in essence executed in raw Lisp. Hence:
- Syntax checking is turned off; and
- Guard checking is completely disabled.
Table events, including logic, are ignored, as are many other events, including defthm and comp.
Soundness claims are weakened for any ACL2 session in which raw mode was ever entered; see defttag.
The normal undoing mechanism (see ubt) is not supported.
Unexpected behavior may occur when you return from raw-mode. For example, if you redefine a :logic mode function whose guards have not been verified, you will not see the change inside the ACL2 loop because there, the raw Common Lisp definition is only executed after guards have been verified; see guards-and-evaluation and see guard-evaluation-table.

We conclude with some details.

Printing results. The rules for printing results are mostly unchanged for raw mode, even to the point of attempting appropriate printing for state and stobjs (though this is merely heuristic when the top-level function is defined in raw mode), for example as follows.

ACL2 !>(defstobj st fld)

Summary
Form:  ( DEFSTOBJ ST ...)
Rules: NIL
Time:  0.02 seconds (prove: 0.00, print: 0.00, other: 0.02)
 ST
ACL2 !>(set-raw-mode-on!)

TTAG NOTE: Adding ttag :RAW-MODE-HACK from the top level loop.
ACL2 P>(defun f (st) st)
F
ACL2 P>(f st)
<st>
ACL2 P>(defun g (st state) (mv 3 st state))
G
ACL2 P>(g st state)
(3 <st> <state>)
ACL2 P>(defun h (state) (mv nil 17 state))
H
ACL2 P>(h state) ; notice the leading space; see below
 17
ACL2 P>

There is however one major exception. If the value to be printed contains any Lisp object that is not a legal ACL2 object, then the print routine is used from the host Lisp, rather than the usual ACL2 printing routine. The following example illustrates the printing used when an illegal ACL2 object needs to be printed. Notice how that ``command conventions'' are observed, as indicated in the ``leading space'' comment above; see ld-post-eval-print. In particular, the ``[Note'' occurs one space over in the second example, and no result is printed in the third example.

ACL2 P>(find-package "ACL2")
[Note:  Printing non-ACL2 result.]
#<The ACL2 package>
ACL2 P>(mv nil (find-package "ACL2") state)
 [Note:  Printing non-ACL2 result.]
#<The ACL2 package>
ACL2 P>(mv t (find-package "ACL2") state)
ACL2 P>(mv 3 (find-package "ACL2"))
[Note:  Printing non-ACL2 result.]
(3 #<The ACL2 package>)
ACL2 P>

If you have trouble with large structures being printed out, you might want to execute appropriate Common Lisp forms in raw mode, for example, (setq *print-length* 5) and (setq *print-level* 5).

Evaluation in raw mode attempts to maintain global stobjs, but may not accommodate arbitrary raw Lisp hacks that hide the stobj returned by ACL2. This can happen in two cases: when array resizing is used on the unique (array) field of a stobj, and when swap-stobjs is used. Consider for example the following log. Evaluation behaves nicely for all forms except the last, where resizing seems not to have taken effect as explained after the log, below.

ACL2 !>(defstobj st fld)

Summary
Form:  ( DEFSTOBJ ST ...)
Rules: NIL
Time:  0.02 seconds (prove: 0.00, print: 0.00, other: 0.02)
 ST
ACL2 !>(defstobj st2 (ar :type (array t (8)) :resizable t))

Summary
Form:  ( DEFSTOBJ ST2 ...)
Rules: NIL
Time:  0.02 seconds (prove: 0.00, print: 0.00, other: 0.02)
 ST2
ACL2 !>(set-raw-mode-on!)

TTAG NOTE: Adding ttag :RAW-MODE-HACK from the top level loop.
ACL2 P>(progn (update-fld 3 st) 17)
17
ACL2 P>(resize-ar 20 st2)
<st2>
ACL2 P>(ar-length st2)
20
ACL2 P>(progn (resize-ar 30 st2) 17)
17
ACL2 P>(ar-length st2) ; Should be 30, but it's not!
20
ACL2 P>

The reason for the last form's ``wrong'' result (and the other ``correct'' results) is technical. A stobj with a single array field is exactly that array in raw Lisp (except in some cases where that may be an array of characters), in which case resizing completely replaces the stobj in the global structure that stores the global stobj values. The use of raw-Lisp progn hides the evidence of stobj modification, which is based on the stobjs returned by the form (but this progn form returns 17). Such evidence is available in the earlier call of resize-ar above, since ACL2 knows that resize-ar returns a st2 stobj. The update-fld call does not provide such evidence for st1, again because of the progn around the call; but update-fld actually modifies the stobj in place, rather than replacing it in the aforementioned global structure.

Include-book. The events add-include-book-dir, add-include-book-dir!, delete-include-book-dir, and delete-include-book-dir! have been designed to work with raw mode. However, if you enter raw mode and then evaluate such forms, then the effects of these forms will disappear when you exit raw mode, in which case you can expect to see a suitable warning. Regarding include-book itself: it should work in raw mode as you might expect, at least if a compiled file or expansion file was created when the book was certified; see certify-book.

Subtopics

Include-raw: A better way to load raw Lisp code than directly using progn! or set-raw-mode.