Istari tactics have the type tactic
; they operate on the current
goal, generating zero or more subgoals.
idtac : tactic
Does nothing.
fail : string -> tactic
Fails, inducing backtracking. Takes a error string that is displayed to the user if it backtracks all the way to the top level.
done : tactic
Like fail
but without the stigma. Intended for marking points where
no subgoals are expected.
cut : tactic -> tactic
Runs the argument tactic, then prunes any backtracking points it creates so subsequent tactics will not backtrack into the argument tactic.
lift : (unit -> tactic) -> tactic
Determines the tactic only when it is time to use it.
andthen : tactic -> tactic -> tactic
The tactic andthen tac1 tac2
runs tac1
, then runs tac2
on all
its subgoals.
A synonym is the infix operator >>
.
andthen1 : tactic -> tactic -> tactic
Like andthen
but generates an error if the first tactic does not
return exactly one subgoal.
A synonym is the infix operator >>+
.
andthenl : tactic -> tactic list -> tactic
The tactic andthenl tac [tac1, ..., tacn]
runs tac
, then runs
tac1
on the first subgoal, tac2
on the second, etc. The initial
tactic tac
must produce exactly n subgoals.
A synonym is the infix operator >>>
.
andthenlPad : tactic -> tactic list -> tactic -> tactic
Like andthenl
except the final tactic argument is run on all
subgoals beyond the nth. The initial tactic must produce at least
n subgoals.
andThenOn : int -> tactic -> tactic -> tactic
The tactic andthenOn n tac1 tac2
runs tac1
, then runs tac2
on
the nth subgoal.
andthenSeq : tactic list -> tactic
Runs each of the tactics in the list in sequence, combined by andthen
.
Similar to foldr andthen idtac
but lazier.
attempt : tactic -> tactic
The tactic attempt tac
applies tac
. If tac
fails, the tactic
does nothing (but succeeds). Equivalent to first [tac, idtac]
.
first : tactic list -> tactic
Applies each tactic in turn, until one succeeds. If none of them succeed, the failure message from the last tactic is used as the failure message.
repeat : tactic -> tactic
The tactic repeat tac
repeatedly applies tac
(recursively on all
subgoals) until it fails.
repeatn : int -> tactic -> tactic
The tactic repeatn n tac
repeatedly applies tac
(recursively on
all subgoals) n times.
orthen : tactic -> (unit -> tactic) -> tactic
The tactic orthen tac1 thunk2
applies tac1
. If it fails, it
calls thunk2
to obtain a second tactic that it then applies.
Equivalent to first [tac1, lift thunk2]
. Designed to work in
conjunction with do
notation.
ifthen : tactic -> tactic -> tactic -> tactic
The tactic ifthen tac1 tac2 tac3
applies tac1
. If it succeeds,
it calls tac2
on its subgoals. If it fails, it calls tac3
. The
combininator commits to its choice: tac2
will not backtrack into
tac1
or tac3
, and tac3
will not backtrack into tac1
. This
distinguishes it from first [andthen tac1 tac2, tac3]
.
Many tactics take a hypothesis or a list of hypotheses as an argument. Hypotheses are given by the grammar:
Hypothesis ::= [name]
[number]
# [number]
\ ... antiquoted name ... \
# \ ... antiquoted number ... \
Hypothesis numbers count backward from the end of the context,
starting with 0. The #
to indicate a number is optional for literal
numbers, but it is required for antiquoted numbers.
In IML hypotheses are given by the datatype:
datatype hypothesis = NAME of symbol | NUMBER of int
which appears in the Hyp
structure.
intro /[ipattern] ... [ipattern]/
Introduces the conclusion type, generating hypotheses. Typically a
pattern will be a simple name, but more expressive patterns are
possible. (See Destruction.) Can introduce
function-like types (forall
, ->
, -t>
, -k>
, intersect
,
iforall
, -g>
, foralltp
), subtypes, and let bindings.
introRaw /[ipattern] ... [ipattern]/
As intro
but does not invoke the typechecker.
intros
Repeatedly introduces the conclusion type, generating names as
needed. Similar to repeat (intro /?/)
.
introsRaw
As intros
but does not invoke the typechecker.
split
Introduces the conclusion type, when doing so requires no choices and generates no hypotheses. Can introduce products, unit, future, and squash.
splitn [n]
Calls split
n
times, continuing into the right-hand subgoal each time.
left
Introduces a sum in the conclusion, selecting the first disjunct.
leftRaw
As left
but does not invoke the typechecker.
right
Introduces a sum in the conclusion, selecting the second disjunct.
rightRaw
As right
but does not invoke the typechecker.
exists /[term]/
Introduces an existential in the conclusion, using the term as the witness.
existsRaw /[term]/
As exists
but does not invoke the typechecker.
existses [/[term]/, ..., /[term]/]
Introduces several existentials in sequence, using the terms as the witnesses.
existsesRaw [/[term]/, ..., /[term]/]
As existses
but does not invoke the typechecker.
exact /[term M]/
Proves the conclusion using M
.
exactRaw /[term M]/
As exact
but does not invoke the typechecker.
unhide
Unhides hidden hypotheses, when the conclusion or any hypotheses are computationally trivial.
unhideRaw
As unhide
but does not invoke the typechecker.
contrapositive /[hyp h]/
If h
’s type is A -> C
and the conclusion is B -> C
, replaces h
with B
and the conclusion with A
. This is particularly useful
when C
is void
; that is, when h
’s type is not A
and the
conclusion is not B
.
contrapositiveRaw /[hyp h]/
As contrapositive
but does not invoke the typechecker.
introOf /[ipattern] ... [ipattern]/
Proves a goal of the form M : A
, where A
is a function type
taking at least the given number of arguments. Typically a pattern
will be a simple name, but more expressive patterns are possible.
(See Destruction.)
introOfRaw /[ipattern] ... [ipattern]/
As introOf
but does not invoke the typechecker.
existsOf /[term M]/
Proves a goal of the form N : A
where A
is either a union
or
iexists
, using M
as the witness.
existsOfRaw /[term M]/
As existsOf
but does not invoke the typechecker.
splitOf
Proves a goal of the form M : A
where A
is either a set
,
iset
, or quotient
. A little more streamlined than the action of
typecheck1
on such a goal.
splitOfRaw
As splitOf
but does not invoke the typechecker.
hyp /[hyp x]/
Proves the current goal if x
’s type matches the conclusion.
assumption
Proves the current goal if any hypothesis’s type matches the conclusion.
hypof
Proves a goal of the form x : A
, if x
’s type is A
.
eassumption
Proves a goal of the form [evar] : A
if there exists a
hypothesis with type A
.
This behavior is usually not desired if the evar is used
elsewhere, so this is not part of the autotactic. It is intended
for situations in which a lemma has just been applied that has a
non-dependent antecedent represented using forall
instead of
->
. (For example, a datatype iterator when the predicate does
not depend on the identity of the iterated term.)
rename /[hyp]/ /[name]/
Renames the indicated hypothesis to use the indicated name.
reintro /[ipattern] ... [ipattern]/
Each pattern must be a name, ?
, or _
. Renames the last
hypotheses in the context so that each hypothesis uses the supplied
name, retains the old name if the ?
pattern is supplied, or is
cleared if the _
pattern is supplied.
renameBefore /[ipattern] ... [ipattern]/ /[hyp x]/
As reintro
except it operates on the last hypotheses that precede x
.
renameBefore /[ipattern] ... [ipattern]/ /concl/
Equivalent to reintro /[ipattern] ... [ipattern]/
.
renameAfter /[ipattern] ... [ipattern]/ /[hyp x]/
As reintro
except it operates on the hypotheses immediately
following x
.
clear /[hyp] ... [hyp]/
Deletes the indicated hypotheses.
weaken [m] [n]
Deletes hypotheses numbered m
through m+n-1
(counting backward from 0).
clearAll
Deletes all hypotheses.
renameOver [hyp x] [hyp y]
Deletes y
and renames x
to y
.
moveBefore /[hyp] ... [hyp]/ /[target hyp]/
Moves the indicated hypotheses immediately before the target hypothesis.
moveBefore /[hyp] ... [hyp]/ /concl/
Moves the indicated hypotheses to the end.
moveBeforeDeps /[hyp] ... [hyp]/ /[target hyp]/
Moves the indicated hypotheses, and any hypotheses they depend on, immediately before the target hypothesis. The moved hypotheses will appear in the same order they appeared in originally.
exchange [m] [n] [p]
Swaps hypotheses numbered m
through m+n-1
with m+n
through
m+n+p-1
(counting backward from 0).
movePos /[hyp] ... [hyp]/ [n]
Moves the indicated hypotheses to position n
(counting backward
from 0).
moveAfter /[hyp] ... [hyp]/ /[target hyp]/
Moves the indicated hypotheses immediately after the target hypothesis.
copy /[hyp]/ /[name]/
Creates a copy of hyp
using the indicated name.
revert /[hyp] ... [hyp]/
Moves the indicated hypotheses into the conclusion.
revertDep /[hyp] ... [hyp]/
As revert
except it always uses forall
, even when an arrow
would suffice.
revert0 [bool]
Moves the last hypothesis into the conclusion. If bool
is true,
then it uses forall
even if an arrow would suffice.
set /[name x]/ /[term M]/
Creates a binding of x = M
.
squashHidden /[hyp] ... [hyp]/
Turns hidden hypotheses of A
into an unhidden hypotheses of
{ A }
. There must be no dependencies on the hypothesis.
reflexivity
Proves a reflexivity goal, such as M = M : A
.
reflexivityRaw
As reflexivity
but does not invoke the typechecker.
symmetry
Applies symmetry to the conclusion. For example, proves M = N : A
generating subgoal N = M : A
.
symmetryRaw
As symmetry
but does not invoke the typechecker.
symmetryIn /[hyp x]/
Applies symmetry to the type of hypothesis x
.
symmetryInRaw /[hyp x]/
As symmetry
but does not invoke the typechecker.
transitivity /[term M]/
Applies symmetry to the conclusion, using M
as the mediating term.
transitivityRaw /[term M]/
As transitivity
but does not invoke the typechecker.
etransitivity
As transitivity
but uses an evar as the mediating term.
compat
Proves a goal of the form h M1 ... Mn = h N1 ... Nn : A
, where h
is a constant or variable, generating subgoals of the form
Mi = Ni : Bi
.
If the head h
takes invisible
arguments, they can be
supplied using ap
just as in typechecking, but only in the first
equand.
compatRaw
As compat
but does not invoke the typechecker.
injection /[hyp x]/
Uses injectivity on an equality in the type of x
. For
example, if x
has type inl M = inl N : A % B
, it
would generate the hypothesis M = N : A
. If the equality is
impossible (e.g., inl M = inr N : A % B
) then it discharges the
goal.
injectionRaw /[hyp x]/
As injection
but does not invoke the typechecker.
decompEq [n] /[term A]/
Proves a goal of the form M ..spine.. = N ..spine.. : B
, where
both spines have length n
, generating the subgoal M = N : A
.
The type B
must be appropriate to A
and spine
.
applyEq /[term F]/ /[term B]/ /[hyp x]/ /[name option y]/
If x
has type M = N : A
and F
has type A -> B
, generates a
hypothesis y
with type F M = F N : B
. (A name is invented if no
name is supplied.) If F
is a path, B
can usually be omitted.
applyEqRaw /[term F]/ /[term B]/ /[hyp x]/ /[name option y]/
As applyEq
but does not invoke the typechecker.
introEq /[name option] ... [name option]/
Proves a goal of the form M = N : A
, decomposing one layer of A
for each argument given, when A
is forall
, ->
, -t>
,
-k>
, or intersect
.
The tactic first establishes M : A
and N : A
and preserves that
fact throughout the process. This can greatly reduce the number of
secondary subgoals. However, if M
, N
, or A
is not already
known to be well-formed, the extensionality
tactic may be better
suited.
introEqRaw /[name option] ... [name option]/
As introEq
but does not invoke the typechecker.
extensionalityAuto
Proves a goal of the form M = N : A
, repeatedly decomposing A
as
long as it is forall
, ->
, -t>
, -k>
, intersect
, exists
,
&
, set
, iset
, or various flavors of unit. The tactic stops
when it reaches an unsupported type, or when it reaches a type
marked manual. (A type is marked manual when it has the form
manuals _
. The other
variants of manual can also be
used, but they are less useful here.)
In addition to automating extensionality, extensionalityAuto
first
establishes M : A
and N : A
and preserves that fact throughout
the process. This can greatly reduce the number of secondary
subgoals. However, if M
, N
, or A
is not already known to be
well-formed, the extensionality
tactic may be better suited.
extensionalityAutoRaw
As extensionalityAuto
but does not invoke the typechecker.
extensionality
Proves a goal of the form M = N : A
, for many primitive types
A
. Unlike introEq
or extensionalityAuto
, it does not
establish M : A
and N : A
first, which may result in fewer or
better subgoals if those typings are not already known.
For equality at union
or iexists
, use existsEq
instead.
extensionalityRaw
As extensionality
but does not invoke the typechecker.
existsEq /[term M]/
Proves a goal of the form N = P : A
where A
is either a union
or iexists
, using M
as the witness.
existsEqRaw /[term M]/
As existsEq
but does not invoke the typechecker.
univIntroEqtype
Proves a goal of the form A = B : U i
, generating subgoals
A = B : type
, A : U i
, and B : U i
.
ofEquands /[hyp h]/ /[name] [name]/
If h
has the type M = N : A
, generates hypotheses with the types
M : A
and N : A
using the supplied names. The second argument
is an intro pattern, so either name can be replaced with _
or ?
.
substitution /[hyp x]/ /[term M]/
Replaces occurrences of x
with M
. Note that M
must not refer
to hypotheses after x
.
substitutionRaw /[hyp x]/ /[term M]/
As substitution
but does not invoke the typechecker.
subst /[hyps]/
For each hypothesis x
in the list: finds another hypothesis with
type x = M : A
or M = x : A
, and replaces occurrences of x
with M
.
substRaw /[hyps]/
As subst
but does not invoke the typechecker.
substStrict /[hyps]/
As subst
but will not move hypotheses to make substitution
possible. Thus when substituting M
for x
, M
must not refer
to hypotheses after x
.
On rare occasions substStrict
might work when subst
fails,
because subst
might fail due to a spurious cyclic dependency
(one that could be eliminated pruning an evar dependency).
substStrictRaw /[hyps]/
As substStrict
but does not invoke the typechecker.
substAll
Carry out all available substitutions, using subst
. Prefers
left-to-right when either direction is possible.
substCautious /[hyps]/
As subst
but will not substitute into the conclusion, and will
never generate a proof obligation that the conclusion is
well-formed.
substCautiousRaw /[hyps]/
As substCautious
but will not invoke the typechecker.
substCautiousStrict /[hyps]/
As substStrict
but will not substitute into the conclusion.
substCautiousStrictRaw /[hyps]/
As substCautiousStrict
but will not invoke the typechecker.
The substitution tactics above first attempt to use a substitution
rule that does not allow the substitution variable (i.e., the
variable being substituted for) to appear in the conclusion. If that
fails, they use a more general rule that allows it. The former rule
is preferable, because it does not generate a proof obligation that
the conclusion is well-formed. However, in unusual circumstances this
can result in a surprising behavior: if the conclusion mentions an
evar that might or might not depend on the substitution variable, that
possible dependency is pruned out. More predictable behavior can be
obtained using substCautious
, which will always prune out any
possible dependency in the conclusion.
subsume /[term A]/
If the conclusion is M : B
, replaces it with M : A
. If the
conclusion is M = N : B
, replaces it with M = N : A
. Otherwise,
if the conclusion is B
, replaces it with A
. In any case,
generates the subgoal A <: B
.
esubsume
As subsume
but uses an evar in place of A
.
eqtp /[term A]/
As subsume
but generates the subgoal A = B : type
.
eeqtp
As eqtp
but uses an evar in place of A
.
forceExact /[term M]/
If M : A
and the conclusion is B
, generates the subgoal A = B :
type
.
forceExactRaw /[term M]/
As forceExact
but does not invoke the typechecker to prove M : A
.
tighten /[hyp x]/ /[term A]/
If x
is bound with type B
, changes its binding to A
, provided
A <: B
and x : A
. Invokes the typechecker on the A <: B
subgoal.
tightenRaw /[hyp x]/ /[term A]/
As tighten
but does not invoke the typechecker.
Destruction and other tactics use intro patterns, defined by the datatype:
datatype ipattern =
Wild
| Ident of Symbol.symbol option
| And of ipattern list
| Or of ipattern list
Wild
discards the term being matched. Ident
creates a hypothesis
for the term being matched, using the indicated name (if SOME
) or
inventing a new name (if NONE
). And
breaks a conjunctive term
into component parts, matching a pattern against each of them. Or
case-analyses a disjunctive term, generating a subgoal for each case,
and matching a pattern against the disjunct in each case.
The syntax for intro patterns is:
PatternAtom ::= 0 (Or [])
_ (Wild)
? (Ident NONE)
[name] (Ident (SOME name))
(Pattern)
[PatternAtom ... PatternAtom] (And [p1, ..., pn])
{PatternArm | ... | PatternArm} (Or [p1, ..., pn])
PatternArm ::= PatternAtom ... PatternAtom (And [p1, ..., pn])
PatternSeq ::= PatternAtom
PatternAtom PatternSeq (And [p1, p2])
Pattern ::= PatternSeq
PatternSeq | Pattern (Or [p1, p2])
[epsilon] (And [])
[epsilon] | Pattern (Or [And [], p])
For example:
Syntax | Parses to: |
---|---|
p1 p2 p3 |
And [p1, And [p2, p3]] |
[p1 p2 p3] |
And [p1, p2, p3] |
p1 p2 p3 | p4 | p5 |
Or [And [p1, And [p2, p3]], Or [p4, p5]] |
{p1 p2 p3 | p4 | p5} |
Or [And [p1, p2, p3], And [p4], And [p5]] |
The effect of the pattern match depends on the type being matched:
Matching on a product, existential type, or set type using p1 p2
(i.e., And [p1, p2]
) will split the term into its two
components.
Matching on a sum using p1 | p2
(i.e., Or [p1, p2]
) will
generate a subgoal for each disjunct.
Matching on void using 0
(i.e., Or []
) will discharge the
goal.
Matching on unit (or other computationally trivial types such as
equality) using ()
(i.e., And []
) will replace the term with
()
.
Matching on future or squash using [p]
(i.e., And [p]
) will
match p
against the underlying term.
Matching on bool using |
(i.e., Or [And [], And []]
) will
generate subgoals for true and false.
Matching on nat using | p
(i.e., Or [And [], p]
) will generate
subgoals for zero and successor.
Matching on a list using | p1 p2
(i.e., Or [And [], And [p1,
p2]]
) will generate subgoals for nil and cons. (This is a special
case of matching on datatypes.)
Matching on a datatype with a sum-of-products pattern {... | ... |
...}
will generate a subgoal for each datatype constructor, and in
each goal it will split the datatype construct into its arguments,
matching on each of them.
In addition, in each goal additional names will be bound to equalities that result from matching the datatype’s index terms against its type.
Matching against a quotient using [p1]
or [p1 p2]
(i.e., And
[p1]
or And [p1, p2]
) will match p1
against a term of the
underlying type, and match p2
(if supplied) against a hypothesis
stating that term is equivalent to itself.
Matching against a quotient using [p1 p2 p3]
(i.e.,
And [p1, p2, p3]
) will match p1
and p2
against two terms of
the underlying type, and match p3
against a hypothesis stating
that they are equivalent. The conclusion must have the form
M : A
, M = N : A
, A : type
, or A = B : type
. The resulting
conclusion will be an equality in which the two equands refer to the
respective equivalent terms.
The destruction tactics are:
destruct /[hyp x]/ /[ipattern]/
Destructs x
using the intro pattern.
destructRaw /[hyp x]/ /[ipattern]/
As destruct
but does not invoke the typechecker.
assert /[term A]/ /[ipattern]/
Generates a subgoal to prove A
, creates a hypothesis of type A
,
and matches the intro pattern against it.
assertRaw /[term A]/ /[ipattern]/
As assert
but does not invoke the typechecker.
assertThen /[term A]/ /[ipattern]/ [tac]
As assert
but then run tac
on the first subgoal.
destructSet /[hyp x]/ /[name]/
If x
has type M : { y : A | B(y) }
, replaces x
by M : A
and
creates a new hypothesis (using the given name) with the type
B(M)
.
destructSetRaw /[hyp x]/ /[name]/
As destructSet
but does not invoke the typechecker.
destructThin /[hyp x]/ /[ipattern]/
Destructs x
, discharging impossible cases and simplifying the
resulting equations. The pattern must be a sum of products (i.e.,
{ ... | ... }
) containing only identifers and ?
.
The tactic can work poorly when (1) x
is mentioned in the
conclusion, (2) destruction generates equations that involve a
variable, and (3) the conclusion is not already known to be
well-formed.
(When destruction results in M
being substituted for x
, and the
simplified equations include y = N : A
where y
is part of M
,
destructThin
invokes substitution to resolve the equation. If x
was mentioned in the original conclusion, then y
will be mentioned
in the post-destruction conclusion, and invoking substitution on a
variable mentioned in the conclusion generates a proof obligation
that the conclusion is well-formed. Often this is fine, but it can
be unfortunate when one cannot yet show that the conclusion is
well-formed, such as when one is proving a typing lemma. This
behavior can be prevented using destructThinCautious
, which will
leave such equations unresolved instead.)
destructThinRaw /[hyp x]/ /[ipattern]/
As destructThin
but does not invoke the typechecker.
destructThinCautious /[hyp x]/ /[ipattern]/
As destructThin
but will not substitute into the conclusion in
the course of resolving index equations, and will never generate a
proof obligation that the conclusion is well-formed. (The
replacement of x
by its various cases still takes place, of
course.)
destructThinCautiousRaw /[hyp x]/ /[ipattern]/
As destructThinCautious
but does not invoke the typechecker.
inversion /[hyp x]/
As destructThin
but it copies x
and destructs the copy. The new
hypothesis being destructed will not appear in the conclusion, no
hypotheses will be disturbed, and any resulting hypotheses will
appear at the bottom.
inversionRaw /[hyp x]/
As inversion
but does not invoke the typechecker.
mimic /[hyp h]/ /[names]/
An ersatz substitution tactic suitable for when the conclusion is
not yet known to be well-formed. If h
’s type is x = M : A
or
M = x : A
where M
’s head is a constructor, the tactic makes x
look like M
by destructing x
and discharging all the
possibilities that don’t look like M
. Will generate new variables
for the subterms of x
, and new equations relating those new
variables to M
’s subterms. The subterm variables will be named
using the supplied names (additional names will be invented if not
enough are supplied) and will appear in the context where x
was.
The equations’ names will be invented, and will appear at the
bottom.
Mimic will never substitute for any variable that appears in the
conclusion. Thus, it is suitable for situations in which the
conclusion is not yet known to be well-formed. If the conclusion
is known to be well-formed, subst
is always preferable.
mimicRaw /[hyp]/ /[names]/
As mimic
but does not invoke the typechecker.
so /[term M]/ /[ipattern]/
If M
has type A
, creates a hypothesis of type A
, and matches
it against the indicated pattern. (See Destruction.)
The term M
can contain placeholders, written with two underscores
(__
). Placeholder subterms result in additional subgoals.
soRaw /[term M]/ /[ipattern]/
As so
but does not invoke the typechecker.
apply /[term M]/
Proves the goal by backchaining through M
. The term may contain
placeholders as in so
. Often M
is just the name of a lemma or
hypothesis.
applyRaw /[term M]/
As apply
but does not invoke the typechecker.
exploit /[term M]/ /[ipattern]/
If M
has type A1 -> ... -> An -> B
, generates subgoals for A1
through An
, and matches the result (of type B
) against the
pattern. The term M
may contain placeholders as in so
.
exploitRaw /[term M]/ /[ipattern]/
As exploit
but does not invoke the typechecker.
eexploit /[term M]/ /[ipattern]/
As exploit but will also proceed through forall
quantifiers,
instantiating them with evars.
eexploitRaw /[term M]/ /[ipattern]/
As eexploit
but does not invoke the typechecker.
witness /[term M]/
Proves the conclusion using M
. Unlike exact
, M
may contain
placeholders as in so
.
witnessRaw /[term M]/
As witness
but does not invoke the typechecker.
generalize /[term M]/ /[term A]/ /[name option x]/
If M : A
, replaces all occurrences of M
in the conclusion with
a new hypothesis x
. (A name is invented if no name is supplied.)
generalizeRaw /[term M]/ /[term A]/ /[name option x]/
As generalize
but does not invoke the typechecker.
generalizeAt /[term M]/ /[term A]/ /[numbers]/ /[name option x]/
As generalize
, but only replaces the indicated appearances of
M
. For example, if [numbers]
is 0 2
then the first and
third appearances of M
are replaced.
generalizeAtRaw /[term M]/ /[term A]/ /[numbers]/ /[name option x]/
As generalizeAt
but does not invoke the typechecker.
remember /[term M]/ /[term A]/ /[name option x]/ /[name option H]/
If M : A
, replaces all occurrences of M
in the conclusion with a
new hypothesis with x
. Also creates H : (x = M : A)
. (Names are
invented if not supplied.)
rememberRaw /[term M]/ /[term A]/ /[name option x]/ /[name option H]/
As remember
but does not invoke the typechecker.
rememberAt /[term M]/ /[term A]/ /[numbers]/ /[name option x]/ /[name option H]/
As remember
, but only replaces the indicated appearances of
M
. For example, if [numbers]
is 0 2
then the first and
third appearances of M
are replaced.
rememberAtRaw /[term M]/ /[term A]/ /[numbers]/ /[name option x]/ /[name option H]/
As rememberAt
but does not invoke the typechecker.
setEq /[name x]/ /[term M]/ /[term A]/ /[name option H]/
If M : A
, creates new hypotheses x : A
and H : (x = M : A)
.
(The name H is invented if not supplied.)
setEqRaw /[name x]/ /[term M]/ /[term A]/ /[name option H]/
As setEq
but does not invoke the typechecker.
boolCase /[term M]/ /[name option H]/
If M : bool
, replaces all occurrences of M
in the conclusion
with a new variable, then splits that variable into true and false
cases. Also creates H : istrue M
and H : not (istrue M)
in the
branches and attempts to rewrite them into a useful form. (The name
H is invented if not supplied.)
boolCaseRaw /[term M]/ /[name option H]/
As boolCase
but does not invoke the typechecker.
boolEq [bool b] /[term M]/
If M : bool
, replaces all occurences of M
in the conclusion with
true
(if b
) or false
(if not b
). Generates the additional
subgoal istrue M
(if b
) or not (istrue M)
(if not b
) and
attempts to rewrite it into a useful form.
boolEqRaw [bool b] /[term M]/
As boolEq
but does not invoke the typechecker.
sinduction /[hyp x]/
Invokes induction on x
. The form of the subgoals generated
depends on x
’s type. This induction tactic is suitable when the
conclusion is not already known to be a well-formed type. For most
types it provides strong induction (a.k.a. course-of-values
induction), and in any case it provides the strongest induction
available for that type. (The strong induction on datatypes employs
the subterm order.)
sinductionRaw /[hyp x]/
As sinduction
but does not invoke the typechecker.
induction /[hyp x]/
Invokes induction on x
by utilizing the iterator for x
’s type.
The form of the subgoals generated thus depends on x
’s type. This
induction tactic generally produces simpler goals than sinduction
,
but it is not suitable unless the typechecker can establish that the
conclusion is a well-formed type. In particular, it is not suitable
for typing lemmas.
inductionRaw /[hyp x]/
As induction
but does not invoke the typechecker.
muUnivInduction /[hyp x]/ /[level i]/
When x
’s type is an inductive type (i.e., a mu
type), invokes
induction on x
. Unlike how sinduction
would behave,
muUnivInduction
assumes that x
’s type belongs to U i
. It
generates a stronger hypothesis, but also a stronger conclusion.
(This is useful only in unusual circumstances.)
muUnivInductionRaw /[hyp x]/ /[level i]/
As muUnivInduction
but does not invoke the typechecker.
typecheck
If the conclusion is a typechecking goal, attempts to prove it.
When the typechecker generates subgoals, the subgoals have
information attached. That information can be displayed using
Prover.detail ();
or using C-c C-d
in the UI.
withTypecheck [tac]
Runs the tactic tac
, then runs the typechecker on all resulting
subgoals.
withTypecheckSnd [tac]
The tactic tac
must have type Typecheck.priority tacticm
.
Runs tac
, then runs the typechecker on all resulting subgoals
that are passed Secondary
.
datatype priority = Primary | Secondary
typecheck1
Runs the typechecker on the current goal for one layer only.
inference
Runs the typechecker on the current goal for its side-effects only. Used to instantiate evars.
infer /[term M]/ /[name option]/
If M
is a path, proves M : A
and creates a new hypothesis of
that type with the given name. (A name is invented if no name is
supplied.)
inferRaw /[term M]/ /[name option]/
As infer
but does not invoke the typechecker on the path’s
arguments.
inferSpine /[hyp]/ /[term N]/ /[name option]/
The hypothesis must have type M : A
and N
must have the form
__ [spine]
. Proves M spine : B
starting from M : A
, and
creates a new hypothesis of that type with the given name. (A
name is invented if no name is supplied.) For example, if h
has
type M : A & B
then inferSpine /h/ /__ #1/ /h'/
will create a
hypothesis h'
with type M #1 : A
.
inferSpineRaw /[hyp]/ /[term N]/ /[name option]/
As inferSpine
but does not invoke the typechecker on the spine’s
arguments.
typecheckLet /[hyp x]/ /[term A]/ /[name option]/
If x
is a let-bound variable that unfolds to M
, runs the
typechecker to prove M : A
, and creates a new hypothesis of type
x : A
with the given name. (A name is invented if no name is
supplied.) If M
is a path, an underscore will usually suffice for
A
.
typechecker : unit -> unit
Runs the typechecker on all of the current goals.
Note that this is not a tactic, so it is invoked typechecker ();
, not typechecker.
trivialize
When the conclusion is computationally trivial, sets the extract to
some standard, closed term, usually ()
. Leaves the goal unchanged
except hidden hypotheses are unhidden.
trivializeRaw
As trivialize
but does not invoke the typechecker.
Typecheck.trace : bool ref
When set to true, the typechecker traces its process.
auto
Attempts to prove the goal using a variety of elementary tactics,
including backchaining through the hypotheses. Will continue on any
subgoals until reaching the default maximum depth (5). If auto
cannot prove the goal completely, it does nothing.
nauto [n]
As auto
but uses n
as the maximum depth.
autoWith /[lemma name] ... [lemma name]/
As auto
but also backchains using the indicated lemmas. If a
“lemma” is a datatype, it backchains with all the datatype’s
constructors.
nautoWith [n] /[lemma name] ... [lemma name]/
Combines nauto
and autoWith
.
nautoWithRaw [n] /[lemma name] ... [lemma name]/
As nautoWith
except that typechecking subgoals are set aside.
The tactic succeds if only typechecking goals remain. If
primary goals would remain, the tactic does nothing.
existsAuto
Similar to auto
but instantiates existentials with evars.
existsAutoRaw
As existsAuto
but does not invoke the typechecker.
existsAutoWith /[lemma name] ... [lemma name]/
As existsAuto
but also backchains using the indicated lemmas.
If a “lemma” is a datatype, it backchains with all the datatype’s
constructors.
existsAutoWithRaw /[lemma name] ... [lemma name]/
As existsAutoWith
but does not invoke the typechecker.
autoTool /[lemma name] ... [lemma name]/
Backchain once, using one of the indicated lemmas. The lemmas are
utilized in the same fashion as auto
. Fails if no lemma applies.
autoToolRaw /[lemma name] ... [lemma name]/
As autoTool
but does not invoke the typechecker.
omega
Solves arithmetic goals using the Omega decision procedure. (William Pugh. The Omega Test: a fast and practical integer programming algorithm for dependence analysis. Communications of the ACM, August 1992.)
omegaRaw
As omega
but does not run the typechecker.
Omega.counterexample : unit -> unit
Prints a counterexample for the last invocation of Omega to fail.
Omega understands the following constants in arithmetic expressions:
nat and integer literals (integer literals also serve as the literals for natural),
linear arithmetic (succ
, plus
, pred
, minus
, plusz
, negz
,
minusz
, succn
, plusn
, predn
, minusn
),
multiplication (times
, timesz
, timesn
) in which at least one operand is
a literal,
minimum and maximum (min
, max
, minz
, maxz
, minn
, maxn
),
conversions between nat
, integer
, and natural
(nat_to_integer
, integer_to_nat
, nat_to_natural
,
natural_to_nat
, natural_to_integer
, integer_to_nat
).
Other expressions are uninterpreted and taken as additional variables.
In propositions Omega understands:
equal and not-equal at nat
, integer
, or natural
,
nat, integer, and natural inequalities (leq
, lt
, leqz
, ltz
,
leqn
, ltn
),
the propositional connectives (prod
, sum
, arrow
, unit
,
void
).
Other propositions are treated as void
in positive positions, and as
unit
in negative positions.
Note that quantified propositions are not understood. For example,
forall x . x <= x
in a positive position (such as the conclusion) is
treated as void
. Consequently it fails with an empty
counterexample, which may be surprising.
Terms must be identical to be taken as the same uninterpreted
variable. This can be surprising, particularly with terms that
mention evars. For example, the terms `length nat L
and
`length E1 L
are taken as two different terms – even though they
display the same way – which will very likely cause Omega to fail.
To lessen the likelihood of this, omega
(but not omegaRaw
) runs
the inference
tactic first to attempt to resolve evars.
Nevertheless, it is possible for evars to leak through inference.
Some forms require Omega to search multiple possibilities. This
includes prod
in a positive position, sum
or arrow
in a negative
position, and any function that is defined by cases (e.g., minus
,
pred
, min
, max
, or integer_to_nat
). Each appearance of such
a form doubles the effective size of the constraint, which will affect
performance. However, multiple occurrences of the same expression
will not double the size multiple times.
Rewriting, reordering, and case analysis are documented on their own pages.
change /[hyp x]/ /[term A]/
Replaces x
’s type with A
, which must be equivalent.
change /concl/ /[term A]/
As above but replaces the conclusion.
exfalso
Replaces the current goal with void
. Also unhides any hidden hypotheses.
Tactic.sideEffect : (unit -> unit) -> tactic
Execute the function argument for its side effects, and do nothing.
displayTac : string -> tactic
Print the string and do nothing.
trustme
Discharges the current goal. Can only be used if unsafe mode has
been activated by running Unsafe.allow ();
.
The tactic type is a special case of the tactic monad type 'a
tacticm
, which is like an ordinary tactic except it passes a value of
type 'a
to each subgoal. Ordinary tactics are then defined:
type tactic = Message.label tacticm
The label contains information that is attached to the subgoal when it is displayed for the user. Most tactics leave it empty.
Most of the tactic combinators actually have more general types than given above, using the tactic monad:
fail : string -> 'a tacticm
cut : 'a tacticm -> 'a tacticm
lift : (unit -> 'a tacticm) -> 'a tacticm
done : 'a tacticm
andthen : 'a tacticm -> 'b tacticm -> 'b tacticm
andthenl : 'a tacticm -> 'b tacticm list -> 'b tacticm
andthenlPad : 'a tacticm -> 'b tacticm list -> 'b tacticm -> 'b tacticm
andthenOn : int -> 'a tacticm -> 'a tacticm -> 'a tacticm
first : 'a tacticm list -> 'a tacticm
orthen : 'a tacticm -> (unit -> 'a tacticm) -> 'a tacticm
ifthen : 'a tacticm -> 'b tacticm -> 'b tacticm -> 'b tacticm
Several of the combinators also have monadic versions that serve as the monad’s unit and various flavors of bind:
idtacM : 'a -> 'a tacticm
andthenM : 'a tacticm -> ('a -> 'b tacticm) -> 'b tacticm
andthenlM : 'a tacticm -> ('a -> 'b tacticm) list -> 'b tacticm
andthenlPadM : 'a tacticm -> ('a -> 'b tacticm) list -> ('a -> 'b tacticm) -> 'b tacticm
andthenOnM : int -> 'a tacticm -> ('a -> 'a tacticm) -> 'a tacticm
ifthenM : 'a tacticm -> ('a -> 'b tacticm) -> 'b tacticm -> 'b tacticm
The monadic operations andthenM
and andthenlM
have the infix
operators >>=
and >>>=
as synonyms.
The most general flavor of bind is:
andthenFoldM :
'a tacticm
-> ('a -> 'b -> 'c tacticm * 'b)
-> ('b -> string option)
-> 'b
-> 'c tacticm
The tactic andthenFoldM tac1 tacfn finish x
first runs tac1
. It
then folds tacfn
left-to-right over the subgoals. Each invocation
is passed (1) the monadic argument of type 'a
that was passed to
that subgoal, and (2) a current value of type 'b
; and returns a 'c
tacticm
(which is used on that subgoal) and a new value of type 'b
that will be used with the next subgoal. The initial 'b
value is
x
. Suppose the final 'b
value is y
. Then finish y
is
evaluated and if the result is SOME msg
, the entire tactic fails
using error message msg
.
Exceptions tend to work clumsily with continuation-passing style. The
tryf
function mediates the interface between them:
exception Tactic.Tryf of string
Tactic.tryf : (unit -> 'a) -> ('a -> 'b tacticm) -> 'b tacticm
The tactic tryf f tac
will call f ()
to obtain x : 'a
, and then
execute tac x
. But if f ()
raises Tryf msg
, the combinator
calls fail msg
instead.
These tactics are primarily used to implement other tactics:
Tactic.replaceJudgement : Judgement.judgement -> tactic
Replaces the judgement
portion of the current goal, leaving the
directory unchanged.
Tactic.replaceHyp : int -> Judgement.hyp -> tactic
Replaces a particular hypothesis (counting backward from 0).
Tactic.replaceConcl : Term.term -> tactic
Replaces the conclusion.
Tactic.withgoal : (goal -> 'a tacticm) -> 'a tacticm
Invokes its argument tactic, passing the current goal to that tactic.
Tactic.withdir : (Directory.directory -> 'a tacticm) -> 'a tacticm
Invokes its argument tactic, passing the current
directory
(the part of the goal used to
relate names to de Bruijn indices) to that tactic.
Tactic.withidir : (Directory.idirectory -> 'a tacticm) -> 'a tacticm
Invokes its argument tactic, passing the current
idirectory
(the part of the goal used to
turn names into de Bruijn indices) to that tactic.
Tactic.withterm : ETerm.eterm -> (Term.term -> 'a tacticm) -> 'a tacticm
Invokes its argument tactic, passing the internal version of its first argument to that tactic.
withterm /[term]/ [tactic function]
.Tactic.withHeadConst : string -> (Constant.constant -> 'a tacticm) -> 'a tacticm
Invokes its argument tactic, passing it the head constant of the current conclusion as its argument. If the conclusion does not have a head constant, it fails using the string argument as the error message. If the head constant is soft and the argument tactic fails, it tries again with the head constant of the conclusion’s unfolding.
Tactic.transformFailure : (string -> string) -> 'a tacticm -> 'a tacticm
Invokes its argument tactic. In the event that tactic fails, it alters the error message using the supplied function.
It is generally a good idea to combine this with a cut. Otherwise
when a subsequent tactic fails it will backtrack through the
transformFailure
and get its error message altered.
Tactic.setFailure : string -> 'a tacticm -> 'a tacticm
As transformFailure
but it simply replaces the error message.
These primitive operations are discussed as part of primitive tactics:
Tactic.refine : Rule.rule -> tactic
Tactic.chdir : Directory.directory -> tactic
Tactic.cast : Judgement.djudgement -> Refine.validation -> tactic
Tactic.execute : judgement -> tactic -> (Refine.validation, string) Sum.sum