Tactics¶
Tactics specify how to transform the proof state of an incomplete proof to eventually generate a complete proof.
Proofs can be developed in two basic ways: In forward reasoning,
the proof begins by proving simple statements that are then combined to prove the
theorem statement as the last step of the proof. With forward reasoning,
for example,
the proof of A /\ B would begin with proofs of A and B, which are
then used to prove A /\ B. Forward reasoning is probably the most common
approach in human-generated proofs.
In backward reasoning, the proof begins with the theorem statement
as the goal, which is then gradually transformed until every subgoal generated
along the way has been proven. In this case, the proof of A /\ B begins
with that formula as the goal. This can be transformed into two subgoals,
A and B, followed by the proofs of A and B. Coq and its tactics
use backward reasoning.
A tactic may fully prove a goal, in which case the goal is removed from the proof state. More commonly, a tactic replaces a goal with one or more subgoals. (We say that a tactic reduces a goal to its subgoals.)
Most tactics require specific elements or preconditions to reduce a goal;
they display error messages if they can't be applied to the goal.
A few tactics, such as auto, don't fail even if the proof state
is unchanged.
Goals are identified by number. The current goal is number
1. Tactics are applied to the current goal by default. (The
default can be changed with the Default Goal Selector
option.) They can
be applied to another goal or to multiple goals with a
goal selector such as 2: tactic.
This chapter describes many of the most common built-in tactics. Built-in tactics can be combined to form tactic expressions, which are described in the Ltac chapter. Since tactic expressions can be used anywhere that a built-in tactic can be used, "tactic" may refer to both built-in tactics and tactic expressions.
Common elements of tactics¶
Reserved keywords¶
The tactics described in this chapter reserve the following keywords:
by using
Thus, these keywords cannot be used as identifiers. It also declares the following character sequences as tokens:
** [= |-
Invocation of tactics¶
A tactic is applied as an ordinary command. It may be preceded by a goal selector (see Section Goal selectors). If no selector is specified, the default selector is used.
-
Option
Default Goal Selector "toplevel_selector"¶ This option controls the default selector, used when no selector is specified when applying a tactic. The initial value is 1, hence the tactics are, by default, applied to the first goal.
Using value
allwill make it so that tactics are, by default, applied to every goal simultaneously. Then, to apply a tactic tac to the first goal only, you can write1:tac.Using value
!enforces that all tactics are used either on a single focused goal or with a local selector (’’strict focusing mode’’).Although other selectors are available, only
all,!or a single natural number are valid default goal selectors.
Bindings¶
Tactics that take a term as an argument may also accept bindings
to instantiate some parameters of the term by name or position.
The general form of a term with bindings is
termtac with bindings where bindings can take two different forms:
In the first form, if an
identis specified, it must be bound in the type oftermand provides the tactic with an instance for the parameter of this name. If anaturalis specified, it refers to then-th non-dependent premise oftermtac.-
Error
No such binder.¶
-
Error
In the second form, the interpretation of the
one_terms depend on which tactic they appear in. Forinduction,destruct,elimandcase, theone_terms provide instances for all the dependent products in the type oftermtacwhile in the case ofapply, or ofconstructorand its variants, only instances for the dependent products that are not bound in the conclusion oftermtacare required.-
Error
Not the right number of missing arguments.¶
-
Error
Intro patterns¶
Intro patterns let you specify the name to assign to variables and hypotheses
introduced by tactics. They also let you split an introduced hypothesis into
multiple hypotheses or subgoals. Common tactics that accept intro patterns
include assert, intros and destruct.
::=*|**|simple_intropatternsimple_intropattern::=simple_intropattern_closed % term0*simple_intropattern_closed::=naming_intropattern|_|or_and_intropattern|equality_intropatternnaming_intropattern::=ident|?|?identor_and_intropattern::=[ intropattern**| ]|( simple_intropattern*, )|( simple_intropattern & simple_intropattern+& )equality_intropattern::=->|<-|[= intropattern* ]Note that the intro pattern syntax varies between tactics.
Most tactics use simple_intropattern in the grammar.
destruct, edestruct, induction,
einduction, case, ecase and the various
inversion tactics use or_and_intropattern, while
intros and eintros use intropattern*.
The eqn: construct in various tactics uses naming_intropattern.
Naming patterns
Use these elementary patterns to specify a name:
ident— use the specified name?— let Coq generate a fresh name_— discard the matched part (unless it is required for another hypothesis)if a disjunction pattern omits a name, such as
[|H2], Coq will choose a name
Splitting patterns
The most common splitting patterns are:
split a hypothesis in the form
A /\ Binto two hypothesesH1: AandH2: Busing the pattern(H1 & H2)or(H1, H2)or[H1 H2]. Example. This also works onA <-> B, which is just a notation representing(A -> B) /\ (B -> A).split a hypothesis in the form
A \/ Binto two subgoals using the pattern[H1|H2]. The first subgoal will have the hypothesisH1: Aand the second subgoal will have the hypothesisH2: B. Examplesplit a hypothesis in either of the forms
A /\ BorA \/ Busing the pattern[].
Patterns can be nested: [[Ha|Hb] H] can be used to split (A \/ B) /\ C.
Note that there is no equivalent to intro patterns for goals. For a goal A /\ B,
use the split tactic to replace the current goal with subgoals A and B.
For a goal A \/ B, use left to replace the current goal with A, or
right to replace the current goal with B.
( simple_intropattern+,) — matches a product over an inductive type with a single constructor. If the number of patterns equals the number of constructor arguments, then it applies the patterns only to the arguments, and( simple_intropattern+, )is equivalent to[simple_intropattern+]. If the number of patterns equals the number of constructor arguments plus the number oflet-ins, the patterns are applied to the arguments andlet-invariables.( simple_intropattern+& )— matches a right-hand nested term that consists of one or more nested binary inductive types such asa1 OP1 a2 OP2 ...(where theOPnare right-associative). (If theOPnare left-associative, additional parentheses will be needed to make the term right-hand nested, such asa1 OP1 (a2 OP2 ...).) The splitting pattern can have more than 2 names, for example(H1 & H2 & H3)matchesA /\ B /\ C. The inductive types must have a single constructor with two parameters. Example[ intropattern*+| ]— splits an inductive type that has multiple constructors such asA \/ Binto multiple subgoals. The number ofintropatterns must be the same as the number of constructors for the matched part.[ intropattern+ ]— splits an inductive type that has a single constructor with multiple parameters such asA /\ Binto multiple hypotheses. Use[H1 [H2 H3]]to matchA /\ B /\ C.[]— splits an inductive type: If the inductive type has multiple constructors, such asA \/ B, create one subgoal for each constructor. If the inductive type has a single constructor with multiple parameters, such asA /\ B, split it into multiple hypotheses.
Equality patterns
These patterns can be used when the hypothesis is an equality:
->— replaces the right-hand side of the hypothesis with the left-hand side of the hypothesis in the conclusion of the goal; the hypothesis is cleared; if the left-hand side of the hypothesis is a variable, it is substituted everywhere in the context and the variable is removed. Example<-— similar to->, but replaces the left-hand side of the hypothesis with the right-hand side of the hypothesis.[= intropattern*, ]— If the product is over an equality type, applies eitherinjectionordiscriminate. Ifinjectionis applicable, the intropattern is used on the hypotheses generated byinjection. If the number of patterns is smaller than the number of hypotheses generated, the pattern?is used to complete the list. Example
Other patterns
*— introduces one or more quantified variables from the result until there are no more quantified variables. Example**— introduces one or more quantified variables or hypotheses from the result until there are no more quantified variables or implications (->).intros **is equivalent tointros. Examplesimple_intropattern_closed % term*— first applies each of the terms with theapply … intactic on the hypothesis to be introduced, then it usessimple_intropattern_closed. Example
Note
A \/ B and A /\ B use infix notation to refer to the inductive
types or and and.
or has multiple constructors (or_introl and or_intror),
while and has a single constructor (conj) with multiple parameters
(A and B).
These are defined in theories/Init/Logic.v. The "where" clauses define the
infix notation for "or" and "and".
Note
intros p+ is not always equivalent to intros p; ... ; intros p
if some of the p are _. In the first form, all erasures are done
at once, while they're done sequentially for each tactic in the second form.
If the second matched term depends on the first matched term and the pattern
for both is _ (i.e., both will be erased), the first intros in the second
form will fail because the second matched term still has the dependency on the first.
Examples:
Example: intro pattern for /\
- Goal forall (A: Prop) (B: Prop), (A /\ B) -> True.
- 1 goal ============================ forall A B : Prop, A /\ B -> True
- intros.
- 1 goal A, B : Prop H : A /\ B ============================ True
- destruct H as (HA & HB).
- 1 goal A, B : Prop HA : A HB : B ============================ True
Example: intro pattern for \/
- Goal forall (A: Prop) (B: Prop), (A \/ B) -> True.
- 1 goal ============================ forall A B : Prop, A \/ B -> True
- intros.
- 1 goal A, B : Prop H : A \/ B ============================ True
- destruct H as [HA|HB]. all: swap 1 2.
- 2 goals A, B : Prop HA : A ============================ True goal 2 is: True 2 goals A, B : Prop HB : B ============================ True goal 2 is: True
Example: -> intro pattern
- Goal forall (x:nat) (y:nat) (z:nat), (x = y) -> (y = z) -> (x = z).
- 1 goal ============================ forall x y z : nat, x = y -> y = z -> x = z
- intros * H.
- 1 goal x, y, z : nat H : x = y ============================ y = z -> x = z
- intros ->.
- 1 goal x, z : nat H : x = z ============================ x = z
Example: [=] intro pattern
The first
intros [=]usesinjectionto strip(S ...)from both sides of the matched equality. The second usesdiscriminateon the contradiction1 = 2(internally represented as(S O) = (S (S O))) to complete the goal.
- Goal forall (n m:nat), (S n) = (S m) -> (S O)=(S (S O)) -> False.
- 1 goal ============================ forall n m : nat, S n = S m -> 1 = 2 -> False
- intros *.
- 1 goal n, m : nat ============================ S n = S m -> 1 = 2 -> False
- intros [= H].
- 1 goal n, m : nat H : n = m ============================ 1 = 2 -> False
- intros [=].
- No more goals.
Example: (A & B & ...) intro pattern
- Parameters (A : Prop) (B: nat -> Prop) (C: Prop).
- A is declared B is declared C is declared
- Goal A /\ (exists x:nat, B x /\ C) -> True.
- 1 goal ============================ A /\ (exists x : nat, B x /\ C) -> True
- intros (a & x & b & c).
- 1 goal a : A x : nat b : B x c : C ============================ True
Example: * intro pattern
- Goal forall (A: Prop) (B: Prop), A -> B.
- 1 goal ============================ forall A B : Prop, A -> B
- intros *.
- 1 goal A, B : Prop ============================ A -> B
Example: ** pattern ("intros **" is equivalent to "intros")
- Goal forall (A: Prop) (B: Prop), A -> B.
- 1 goal ============================ forall A B : Prop, A -> B
- intros **.
- 1 goal A, B : Prop H : A ============================ B
Example: compound intro pattern
- Goal forall A B C:Prop, A \/ B /\ C -> (A -> C) -> C.
- 1 goal ============================ forall A B C : Prop, A \/ B /\ C -> (A -> C) -> C
- intros * [a | (_,c)] f.
- 2 goals A, B, C : Prop a : A f : A -> C ============================ C goal 2 is: C
- all: swap 1 2.
- 2 goals A, B, C : Prop c : C f : A -> C ============================ C goal 2 is: C
Example: combined intro pattern using [=] -> and %
- Require Import Coq.Lists.List.
- Section IntroPatterns.
- Variables (A : Type) (xs ys : list A).
- A is declared xs is declared ys is declared
- Example ThreeIntroPatternsCombined : S (length ys) = 1 -> xs ++ ys = xs.
- 1 goal A : Type xs, ys : list A ============================ S (length ys) = 1 -> xs ++ ys = xs
- intros [=->%length_zero_iff_nil].
- 1 goal A : Type xs : list A ============================ xs ++ nil = xs
introswould addH : S (length ys) = 1
intros [=]would additionally applyinjectiontoHto yieldH0 : length ys = 0
intros [=->%length_zero_iff_nil]applies the theorem, making H the equalityl=nil, which is then applied as for->.Theorem length_zero_iff_nil (l : list A): length l = 0 <-> l=nil.The example is based on Tej Chajed's coq-tricks
Occurrence clauses¶
An occurrence is a subterm of a goal or hypothesis that matches a pattern provided by a tactic. Occurrence clauses select a subset of the ocurrences in a goal and/or in one or more of its hypotheses.
occurrences::=at occs_nums|in goal_occurrencessimple_occurrences::=occurrencesoccs_nums::=-? nat_or_var+nat_or_var::=naturalidentgoal_occurrences::=hyp_occs+, |- concl_occs??|* |- concl_occs?||- concl_occs?|concl_occs?hyp_occs::=hypident at occs_nums?hypident::=ident|( type of ident )|( value of ident )concl_occs::=* at occs_nums?
occurrencesThe first form of
occurrencesselects occurrences in the conclusion of the goal. The second form can select occurrences in the goal conclusion and in one or more hypotheses.simple_occurrencesA semantically restricted form of
occurrencesthat doesn't allow theatclause anywhere within it.-? nat_or_var+Selects the specified occurrences within a single goal or hypothesis. Occurrences are numbered starting with 1 following a depth-first traversal of the term's expression, including occurrences in implicit arguments and coercions that are not displayed by default. (Set the
Printing Allflag to show those in the printed term.)For example, when matching the pattern
_ + _in the term(a + b) + c, occurrence 1 is(...) + cand occurrence 2 is(a + b). When matching that pattern with terma + (b + c), occurrence 1 isa + (...)and occurrence 2 isb + c.Specifying
-includes all occurrences except the ones listed.hyp_occs*, |- concl_occs??Selects occurrences in the specified hypotheses and the specified occurrences in the conclusion.
* |- concl_occs?Selects all occurrences in all hypotheses and the specified occurrences in the conclusion.
|- concl_occs?Selects the specified occurrences in the conclusion.
goal_occurrences ::= concl_occs?Selects all occurrences in all hypotheses and in the specified occurrences in the conclusion.
hypident at occs_nums?Omiting
occs_numsselects all occurrences within the hypothesis.hypident ::= identSelects the hypothesis named
ident.( type of ident )Selects the type part of the named hypothesis (e.g.
: nat).( value of ident )Selects the value part of the named hypothesis (e.g.
:= 1).concl_occs ::= * at occs_nums?Selects occurrences in the conclusion. '*' by itself selects all occurrences.
occs_numsselects the specified occurrences.Use
in *to select all occurrences in all hypotheses and the conclusion, which is equivalent toin * |- *. Use* |-to select all occurrences in all hypotheses.When rewriting in multiple hypotheses, they must not appear in the term to rewrite. For instance
rewrite H in H,H'is an error. If an hypothesis appears only through a hole, it will be removed from that hole's context.With
rewrite term in *, hypotheses on which the dependency cannot be avoided are skipped, for instancerewrite H in *skips rewriting inH. This is the case even if only one hypothesis ends up rewritten.If multiple occurrences are given, such as in
rewriteH at 1 2 3, the tactic must match at least one occurrence in order to succeed. The tactic will fail if no occurrences match. Occurrence numbers that are out of range (e.g.at 1 3when there are only 2 occurrences in the hypothesis or conclusion) are ignored.Tactics that use occurrence clauses include
set,remember,inductionanddestruct.
Applying theorems¶
-
Tactic
exact term¶ This tactic applies to any goal. It gives directly the exact proof term of the goal. Let
Tbe our goal, letpbe a term of typeUthenexact psucceeds iffTandUare convertible (see Conversion rules).-
Error
Not an exact proof.¶
-
Error
-
Tactic
assumption¶ This tactic looks in the local context for a hypothesis whose type is convertible to the goal. If it is the case, the subgoal is proved. Otherwise, it fails.
-
Error
No such assumption.¶
-
Variant
eassumption¶ This tactic behaves like
assumptionbut is able to handle goals with existential variables.
-
Error
-
Tactic
refine term¶ This tactic applies to any goal. It behaves like
exactwith a big difference: the user can leave some holes (denoted by_or(_ : type)) in the term.refinewill generate as many subgoals as there are remaining holes in the elaborated term. The type of holes must be either synthesized by the system or declared by an explicit cast like(_ : nat -> Prop). Any subgoal that occurs in other subgoals is automatically shelved, as if callingshelve_unifiable. The produced subgoals (shelved or not) are not candidates for typeclass resolution, even if they have a type-class type as conclusion, letting the user control when and how typeclass resolution is launched on them. This low-level tactic can be useful to advanced users.Example
- Inductive Option : Set := | Fail : Option | Ok : bool -> Option.
- Option is defined Option_rect is defined Option_ind is defined Option_rec is defined Option_sind is defined
- Definition get : forall x:Option, x <> Fail -> bool.
- 1 goal ============================ forall x : Option, x <> Fail -> bool
- refine (fun x:Option => match x return x <> Fail -> bool with | Fail => _ | Ok b => fun _ => b end).
- 1 goal x : Option ============================ Fail <> Fail -> bool
- intros; absurd (Fail = Fail); trivial.
- No more goals.
- Defined.
-
Error
Refine passed ill-formed term.¶ The term you gave is not a valid proof (not easy to debug in general). This message may also occur in higher-level tactics that call
refineinternally.
-
Error
Cannot infer a term for this placeholder.¶ There is a hole in the term you gave whose type cannot be inferred. Put a cast around it.
-
Variant
simple refine term¶ This tactic behaves like refine, but it does not shelve any subgoal. It does not perform any beta-reduction either.
-
Variant
notypeclasses refine term¶ This tactic behaves like
refineexcept it performs type checking without resolution of typeclasses.
-
Variant
simple notypeclasses refine term¶ This tactic behaves like the combination of
simple refineandnotypeclasses refine: it performs type checking without resolution of typeclasses, does not perform beta reductions or shelve the subgoals.
Debug"unification"enables printing traces of unification steps used during elaboration/typechecking and therefinetactic."ho-unification"prints information about higher order heuristics.
-
Tactic
apply term¶ This tactic applies to any goal. The argument term is a term well-formed in the local context. The tactic
applytries to match the current goal against the conclusion of the type ofterm. If it succeeds, then the tactic returns as many subgoals as the number of non-dependent premises of the type of term. If the conclusion of the type oftermdoes not match the goal and the conclusion is an inductive type isomorphic to a tuple type, then each component of the tuple is recursively matched to the goal in the left-to-right order.The tactic
applyrelies on first-order unification with dependent types unless the conclusion of the type oftermis of the formP (t1 ... tn)withPto be instantiated. In the latter case, the behavior depends on the form of the goal. If the goal is of the form(fun x => Q) u1 ... unand thetianduiunify, thenPis taken to be(fun x => Q). Otherwise,applytries to definePby abstracting overt_1 ... t__nin the goal. Seepatternto transform the goal so that it gets the form(fun x => Q) u1 ... un.-
Error
Unable to unify term with term.¶ The
applytactic failed to match the conclusion oftermand the current goal. You can help theapplytactic by transforming your goal with thechangeorpatterntactics.
-
Error
Unable to find an instance for the variables ident+.¶ This occurs when some instantiations of the premises of
termare not deducible from the unification. This is the case, for instance, when you want to apply a transitivity property. In this case, you have to use one of the variants below:
-
Variant
apply term with term+ Provides apply with explicit instantiations for all dependent premises of the type of term that do not occur in the conclusion and consequently cannot be found by unification. Notice that the collection
term+must be given according to the order of these dependent premises of the type of term.-
Error
Not the right number of missing arguments.¶
-
Error
-
Variant
apply term with bindings This also provides apply with values for instantiating premises. Here, variables are referred by names and non-dependent products by increasing numbers (see Bindings).
-
Variant
apply term+, This is a shortcut for
apply term1; [.. | ... ; [ .. | apply termn] ... ], i.e. for the successive applications oftermi+1 on the last subgoal generated byapply termi, starting from the application ofterm1.
-
Variant
eapply term¶ The tactic
eapplybehaves likeapplybut it does not fail when no instantiations are deducible for some variables in the premises. Rather, it turns these variables into existential variables which are variables still to instantiate (see Existential variables). The instantiation is intended to be found later in the proof.
-
Variant
rapply term¶ The tactic
rapplybehaves likeeapplybut it uses the proof engine ofrefinefor dealing with existential variables, holes, and conversion problems. This may result in slightly different behavior regarding which conversion problems are solvable. However, likeapplybut unlikeeapply,rapplywill fail if there are any holes which remain intermitself after typechecking and typeclass resolution but before unification with the goal. More technically,termis first parsed as aconstrrather than as auconstroropen_constrbefore being applied to the goal. Note thatrapplyprefers to instantiate as many hypotheses oftermas possible. As a result, if it is possible to applytermto arbitrarily many arguments without getting a type error,rapplywill loop.Note that you need to
Require Import Coq.Program.Tacticsto make use ofrapply.
-
Variant
simple apply term. This behaves like
applybut it reasons modulo conversion only on subterms that contain no variables to instantiate. For instance, the following example does not succeed because it would require the conversion ofid ?fooandO.Example
- Definition id (x : nat) := x.
- id is defined
- Parameter H : forall x y, id x = y.
- H is declared
- Goal O = O.
- 1 goal ============================ 0 = 0
- Fail simple apply H.
- The command has indeed failed with message: Unable to unify "id ?M150 = ?M151" with "0 = 0".
Because it reasons modulo a limited amount of conversion,
simple applyfails quicker thanapplyand it is then well-suited for uses in user-defined tactics that backtrack often. Moreover, it does not traverse tuples asapplydoes.
-
Variant
simple? apply term with bindings?+,¶ -
Variant
simple? eapply term with bindings?+,¶ This summarizes the different syntaxes for
applyandeapply.
-
Variant
lapply term¶ This tactic applies to any goal, say
G. The argument term has to be well-formed in the current context, its type being reducible to a non-dependent productA -> BwithBpossibly containing products. Then it generates two subgoalsB->GandA. Applyinglapply H(whereHhas typeA->BandBdoes not start with a product) does the same as giving the sequencecut B. 2:apply H.wherecutis described below.
-
Error
Example
Assume we have a transitive relation R on nat:
- Parameter R : nat -> nat -> Prop.
- R is declared
- Axiom Rtrans : forall x y z:nat, R x y -> R y z -> R x z.
- Rtrans is declared
- Parameters n m p : nat.
- n is declared m is declared p is declared
- Axiom Rnm : R n m.
- Rnm is declared
- Axiom Rmp : R m p.
- Rmp is declared
Consider the goal (R n p) provable using the transitivity of R:
- Goal R n p.
- 1 goal ============================ R n p
The direct application of Rtrans with apply fails because no value
for y in Rtrans is found by apply:
- apply Rtrans.
- Toplevel input, characters 6-12: > apply Rtrans. > ^^^^^^ Error: Unable to find an instance for the variable y.
A solution is to apply (Rtrans n m p) or (Rtrans n m).
- apply (Rtrans n m p).
- 2 goals ============================ R n m goal 2 is: R m p
Note that n can be inferred from the goal, so the following would work
too.
- apply (Rtrans _ m).
- 2 goals ============================ R n m goal 2 is: R m p
More elegantly, apply Rtrans with (y:=m) allows only mentioning the
unknown m:
- apply Rtrans with (y := m).
- 2 goals ============================ R n m goal 2 is: R m p
Another solution is to mention the proof of (R x y) in Rtrans
- apply Rtrans with (1 := Rnm).
- 1 goal ============================ R m p
... or the proof of (R y z).
- apply Rtrans with (2 := Rmp).
- 1 goal ============================ R n m
On the opposite, one can use eapply which postpones the problem of
finding m. Then one can apply the hypotheses Rnm and Rmp. This
instantiates the existential variable and completes the proof.
- eapply Rtrans.
- 2 focused goals (shelved: 1) ============================ R n ?y goal 2 is: R ?y p
- apply Rnm.
- 1 goal ============================ R m p
- apply Rmp.
- No more goals.
Note
When the conclusion of the type of the term to apply is an inductive
type isomorphic to a tuple type and apply looks recursively whether a
component of the tuple matches the goal, it excludes components whose
statement would result in applying an universal lemma of the form
forall A, ... -> A. Excluding this kind of lemma can be avoided by
setting the following flag:
-
Tactic
apply term in ident¶ This tactic applies to any goal. The argument
termis a term well-formed in the local context and the argumentidentis an hypothesis of the context. The tacticapply term in identtries to match the conclusion of the type ofidentagainst a non-dependent premise of the type ofterm, trying them from right to left. If it succeeds, the statement of hypothesisidentis replaced by the conclusion of the type ofterm. The tactic also returns as many subgoals as the number of other non-dependent premises in the type oftermand of the non-dependent premises of the type ofident. If the conclusion of the type oftermdoes not match the goal and the conclusion is an inductive type isomorphic to a tuple type, then the tuple is (recursively) decomposed and the first component of the tuple of which a non-dependent premise matches the conclusion of the type ofident. Tuples are decomposed in a width-first left-to-right order (for instance if the type ofH1isA <-> Band the type ofH2isAthenapply H1 in H2transforms the type ofH2intoB). The tacticapplyrelies on first-order pattern matching with dependent types.-
Error
Statement without assumptions.¶ This happens if the type of
termhas no non-dependent premise.
-
Error
Unable to apply.¶ This happens if the conclusion of
identdoes not match any of the non-dependent premises of the type ofterm.
-
Variant
apply term with bindings+, in ident+, This does the same but uses the bindings to instantiate parameters of
term(see Bindings).
-
Variant
eapply term with bindings?+, in ident+, This works as
apply … inbut turns unresolved bindings into existential variables, if any, instead of failing.
-
Variant
apply term with bindings?+, in ident as simple_intropattern?+,¶ This works as
apply … inbut applying an associatedsimple_intropatternto each hypothesisidentthat comes with such clause.
-
Variant
simple apply term in ident+, This behaves like
apply … inbut it reasons modulo conversion only on subterms that contain no variables to instantiate and does not traverse tuples. See the corresponding example.
-
Error
Debug "tactic-unification" enables printing traces of
unification steps in tactic unification. Tactic unification is used in
tactics such as apply and rewrite.
Managing the local context¶
-
Tactic
intro¶ This tactic applies to a goal that is either a product or starts with a let-binder. If the goal is a product, the tactic implements the "Lam" rule given in Typing rules 1. If the goal starts with a let-binder, then the tactic implements a mix of the "Let" and "Conv".
If the current goal is a dependent product
forall x:T, U(resplet x:=t in U) thenintroputsx:T(respx:=t) in the local context. The new subgoal isU.If the goal is a non-dependent product \(T \rightarrow U\), then it puts in the local context either
Hn:T(ifTis of typeSetorProp) orXn:T(if the type ofTisType). The optional indexnis such thatHnorXnis a fresh identifier. In both cases, the new subgoal isU.If the goal is an existential variable,
introforces the resolution of the existential variable into a dependent product \(\forall\)x:?X, ?Y, putsx:?Xin the local context and leaves?Yas a new subgoal allowed to depend onx.The tactic
introapplies the tactichnfuntilintrocan be applied or the goal is not head-reducible.-
Error
No product even after head-reduction.¶
-
Variant
intro ident This applies
introbut forcesidentto be the name of the introduced hypothesis.
Note
If a name used by intro hides the base name of a global constant then the latter can still be referred to by a qualified name (see Qualified identifiers).
-
Variant
intros¶ This repeats
intrountil it meets the head-constant. It never reduces head-constants and it never fails.
-
Variant
intros until ident This repeats intro until it meets a premise of the goal having the form
(ident : type)and discharges the variable namedidentof the current goal.-
Error
No such hypothesis in current goal.¶
-
Error
-
Variant
intros until natural This repeats
intrountil thenatural-th non-dependent product.Example
On the subgoal
forall x y : nat, x = y -> y = xthe tacticintros until 1is equivalent tointros x y H, asx = y -> y = xis the first non-dependent product.On the subgoal
forall x y z : nat, x = y -> y = xthe tacticintros until 1is equivalent tointros x y zas the product onzcan be rewritten as a non-dependent product:forall x y : nat, nat -> x = y -> y = x.
-
Variant
intro ident1? after ident2 -
Variant
intro ident1? before ident2 -
Variant
intro ident1? at top -
Variant
intro ident1? at bottom These tactics apply
intro ident1?and move the freshly introduced hypothesis respectively after the hypothesisident2, before the hypothesisident2, at the top of the local context, or at the bottom of the local context. All hypotheses on which the new hypothesis depends are moved too so as to respect the order of dependencies between hypotheses. It is equivalent tointro ident1?followed by the appropriate call tomove,move … before …,move … at top, ormove … at bottom.Note
intro at bottomis a synonym forintrowith no argument.
-
Error
-
Tactic
intros intropattern*¶ Introduces one or more variables or hypotheses from the goal by matching the intro patterns. See the description in Intro patterns.
-
Tactic
eintros intropattern*¶ Works just like
intros …except that it creates existential variables for any unresolved variables rather than failing.
-
Tactic
clear ident¶ This tactic erases the hypothesis named
identin the local context of the current goal. As a consequence,identis no more displayed and no more usable in the proof development.-
Error
No such hypothesis.¶
-
Variant
clear - ident+ This variant clears all the hypotheses except the ones depending in the hypotheses named
ident+and in the goal.
-
Variant
clear This variants clears all the hypotheses except the ones the goal depends on.
-
Error
-
Tactic
revert ident+¶ This applies to any goal with variables
ident+. It moves the hypotheses (possibly defined) to the goal, if this respects dependencies. This tactic is the inverse ofintro.-
Error
No such hypothesis.¶
-
Error
-
Tactic
move ident1 after ident2¶ This moves the hypothesis named
ident1in the local context after the hypothesis namedident2, where “after” is in reference to the direction of the move. The proof term is not changed.If
ident1comes beforeident2in the order of dependencies, then all the hypotheses betweenident1andident2that (possibly indirectly) depend onident1are moved too, and all of them are thus moved afterident2in the order of dependencies.If
ident1comes afterident2in the order of dependencies, then all the hypotheses betweenident1andident2that (possibly indirectly) occur in the type ofident1are moved too, and all of them are thus moved beforeident2in the order of dependencies.-
Variant
move ident1 before ident2¶ This moves
ident1towards and just before the hypothesis namedident2. As formove, dependencies overident1(whenident1comes beforeident2in the order of dependencies) or in the type ofident1(whenident1comes afterident2in the order of dependencies) are moved too.
-
Variant
move ident at top¶ This moves
identat the top of the local context (at the beginning of the context).
-
Variant
move ident at bottom¶ This moves
identat the bottom of the local context (at the end of the context).
-
Error
No such hypothesis.¶
Example
- Goal forall x :nat, x = 0 -> forall z y:nat, y=y-> 0=x.
- 1 goal ============================ forall x : nat, x = 0 -> nat -> forall y : nat, y = y -> 0 = x
- intros x H z y H0.
- 1 goal x : nat H : x = 0 z, y : nat H0 : y = y ============================ 0 = x
- move x after H0.
- 1 goal z, y : nat H0 : y = y x : nat H : x = 0 ============================ 0 = x
- Undo.
- 1 goal x : nat H : x = 0 z, y : nat H0 : y = y ============================ 0 = x
- move x before H0.
- 1 goal z, y, x : nat H : x = 0 H0 : y = y ============================ 0 = x
- Undo.
- 1 goal x : nat H : x = 0 z, y : nat H0 : y = y ============================ 0 = x
- move H0 after H.
- 1 goal x, y : nat H0 : y = y H : x = 0 z : nat ============================ 0 = x
- Undo.
- 1 goal x : nat H : x = 0 z, y : nat H0 : y = y ============================ 0 = x
- move H0 before H.
- 1 goal x : nat H : x = 0 y : nat H0 : y = y z : nat ============================ 0 = x
-
Variant
-
Tactic
rename ident1 into ident2¶ This renames hypothesis
ident1intoident2in the current context. The name of the hypothesis in the proof-term, however, is left unchanged.-
Variant
rename identi into identj+, This renames the variables
identiintoidentjin parallel. In particular, the target identifiers may contain identifiers that exist in the source context, as long as the latter are also renamed by the same tactic.
-
Error
No such hypothesis.¶
-
Variant
-
Tactic
set (ident := term)¶ This replaces
termbyidentin the conclusion of the current goal and adds the new definitionident := termto the local context.If
termhas holes (i.e. subexpressions of the form “_”), the tactic first checks that all subterms matching the pattern are compatible before doing the replacement using the leftmost subterm matching the pattern.-
Variant
set (ident := term) in goal_occurrences This notation allows specifying which occurrences of
termhave to be substituted in the context. Thein goal_occurrencesclause is an occurrence clause whose syntax and behavior are described in goal occurrences.
-
Variant
set (ident binder* := term) in goal_occurrences? This is equivalent to
set (ident := fun binder* => term) in goal_occurrences?.
-
Variant
set term in goal_occurrences? This behaves as
set (ident := term) in goal_occurrences?butidentis generated by Coq.
-
Variant
eset (ident binder* := term) in goal_occurrences?¶ -
Variant
eset term in goal_occurrences?¶ While the different variants of
setexpect that no existential variables are generated by the tactic,esetremoves this constraint. In practice, this is relevant only whenesetis used as a synonym ofepose, i.e. when thetermdoes not occur in the goal.
-
Variant
-
Tactic
remember term as ident1 eqn:naming_intropattern?¶ This behaves as
set (ident := term) in *, using a logical (Leibniz’s) equality instead of a local definition. Usenaming_intropatternto name or split up the new equation.-
Variant
remember term as ident1 eqn:naming_intropattern? in goal_occurrences This is a more general form of
rememberthat remembers the occurrences oftermspecified by an occurrence set.
-
Variant
eremember term as ident1 eqn:naming_intropattern? in goal_occurrences?¶ While the different variants of
rememberexpect that no existential variables are generated by the tactic,erememberremoves this constraint.
-
Variant
-
Tactic
pose (ident := term)¶ This adds the local definition
ident := termto the current context without performing any replacement in the goal or in the hypotheses. It is equivalent toset (ident := term) in |-.
-
Tactic
decompose [qualid+] term¶ This tactic recursively decomposes a complex proposition in order to obtain atomic ones.
Example
- Goal forall A B C:Prop, A /\ B /\ C \/ B /\ C \/ C /\ A -> C.
- 1 goal ============================ forall A B C : Prop, A /\ B /\ C \/ B /\ C \/ C /\ A -> C
- intros A B C H; decompose [and or] H.
- 3 goals A, B, C : Prop H : A /\ B /\ C \/ B /\ C \/ C /\ A H1 : A H0 : B H3 : C ============================ C goal 2 is: C goal 3 is: C
- all: assumption.
- No more goals.
- Qed.
Note
decomposedoes not work on right-hand sides of implications or products.-
Variant
decompose sum term This decomposes sum types (like
or).
Controlling the proof flow¶
-
Tactic
assert (ident : type)¶ This tactic applies to any goal.
assert (H : U)adds a new hypothesis of nameHassertingUto the current goal and opens a new subgoalU2. The subgoalUcomes first in the list of subgoals remaining to prove.-
Error
Not a proposition or a type.¶ Arises when the argument
typeis neither of typeProp,SetnorType.
-
Variant
assert type by tactic This tactic behaves like
assertbut applies tactic to solve the subgoals generated by assert.-
Error
Proof is not complete.¶
-
Error
-
Variant
assert type as simple_intropattern If
simple_intropatternis an intro pattern (see Intro patterns), the hypothesis is named after this introduction pattern (in particular, ifsimple_intropatternisident, the tactic behaves likeassert (ident : type)). Ifsimple_intropatternis an action introduction pattern, the tactic behaves likeassert typefollowed by the action done by this introduction pattern.
-
Variant
assert type as simple_intropattern by tactic This combines the two previous variants of
assert.
-
Error
-
Variant
eassert type as simple_intropattern by tactic¶ While the different variants of
assertexpect that no existential variables are generated by the tactic,eassertremoves this constraint. This lets you avoid specifying the asserted statement completely before starting to prove it.
-
Variant
pose proof term as simple_intropattern?¶ This tactic behaves like
assert type as simple_intropattern? by exact termwheretypeis the type ofterm. In particular,pose proof term as identbehaves asassert (ident := term)andpose proof term as simple_intropatternis the same as applying thesimple_intropatterntoterm.
-
Variant
epose proof term as simple_intropattern?¶ While
pose proofexpects that no existential variables are generated by the tactic,epose proofremoves this constraint.
-
Variant
pose proof (ident := term) This is an alternative syntax for
assert (ident := term)andpose proof term as ident, following the model ofpose (ident := term)but dropping the value ofident.
-
Variant
epose proof (ident := term) This is an alternative syntax for
eassert (ident := term)andepose proof term as ident, following the model ofepose (ident := term)but dropping the value ofident.
-
Variant
enough (ident : type)¶ This adds a new hypothesis of name
identassertingtypeto the goal the tacticenoughis applied to. A new subgoal statingtypeis inserted after the initial goal rather than before it asassertwould do.
-
Variant
enough type This behaves like
enough (ident : type)with the nameidentof the hypothesis generated by Coq.
-
Variant
enough type as simple_intropattern This behaves like
enough typeusingsimple_intropatternto name or destruct the new hypothesis.
-
Variant
enough (ident : type) by tactic -
Variant
enough type as simple_intropattern? by tactic This behaves as above but with
tacticexpected to solve the initial goal after the extra assumptiontypeis added and possibly destructed. If theas simple_intropatternclause generates more than one subgoal,tacticis applied to all of them.
-
Variant
eenough type as simple_intropattern? by tactic?¶ -
Variant
eenough (ident : type) by tactic?¶ While the different variants of
enoughexpect that no existential variables are generated by the tactic,eenoughremoves this constraint.
-
Variant
cut type¶ This tactic applies to any goal. It implements the non-dependent case of the “App” rule given in Typing rules. (This is Modus Ponens inference rule.)
cut Utransforms the current goalTinto the two following subgoals:U -> TandU. The subgoalU -> Tcomes first in the list of remaining subgoal to prove.
-
Variant
specialize (ident term*) as simple_intropattern?¶ -
Variant
specialize ident with bindings as simple_intropattern?¶ This tactic works on local hypothesis
ident. The premises of this hypothesis (either universal quantifications or non-dependent implications) are instantiated by concrete terms coming either from argumentsterm*or from Bindings. In the first form the application toterm*can be partial. The first form is equivalent toassert (ident := ident term*). In the second form, instantiation elements can also be partial. In this case the uninstantiated arguments are inferred by unification if possible or left quantified in the hypothesis otherwise. With theasclause, the local hypothesisidentis left unchanged and instead, the modified hypothesis is introduced as specified by thesimple_intropattern. The nameidentcan also refer to a global lemma or hypothesis. In this case, for compatibility reasons, the behavior ofspecializeis close to that ofgeneralize: the instantiated statement becomes an additional premise of the goal. Theasclause is especially useful in this case to immediately introduce the instantiated statement as a local hypothesis.
-
Tactic
generalize term¶ This tactic applies to any goal. It generalizes the conclusion with respect to some term.
Example
- Goal forall x y:nat, 0 <= x + y + y.
- 1 goal ============================ forall x y : nat, 0 <= x + y + y
- Proof. intros *.
- 1 goal x, y : nat ============================ 0 <= x + y + y
- Show.
- 1 goal x, y : nat ============================ 0 <= x + y + y
- generalize (x + y + y).
- 1 goal x, y : nat ============================ forall n : nat, 0 <= n
If the goal is G and t is a subterm of type T in the goal,
then generalize t replaces the goal by forall (x:T), G′ where G′
is obtained from G by replacing all occurrences of t by x. The
name of the variable (here n) is chosen based on T.
-
Variant
generalize term+ This is equivalent to
generalize term; ... ; generalize term. Note that the sequence of term i 's are processed from n to 1.
-
Variant
generalize term at natural+ This is equivalent to
generalize termbut it generalizes only over the specified occurrences ofterm(counting from left to right on the expression printed using thePrinting Allflag).
-
Variant
generalize term as ident This is equivalent to
generalize termbut it usesidentto name the generalized hypothesis.
-
Variant
generalize term at natural+ as ident+, This is the most general form of
generalizethat combines the previous behaviors.
-
Variant
generalize dependent term This generalizes term but also all hypotheses that depend on
term. It clears the generalized hypotheses.
-
Tactic
evar (ident : term)¶ The
evartactic creates a new local definition namedidentwith typetermin the context. The body of this binding is a fresh existential variable.
-
Tactic
instantiate (ident := term )¶ The instantiate tactic refines (see
refine) an existential variableidentwith the termterm. It is equivalent toonly [ident]: refine term(preferred alternative).Note
To be able to refer to an existential variable by name, the user must have given the name explicitly (see Existential variables).
Note
When you are referring to hypotheses which you did not name explicitly, be aware that Coq may make a different decision on how to name the variable in the current goal and in the context of the existential variable. This can lead to surprising behaviors.
-
Variant
instantiate (natural := term) This variant selects an existential variable by its position. The
naturalargument is the position of the existential variable from right to left in the conclusion of the goal. (Use one of the variants below to select an existential variable in a hypothesis.) Counting starts at 1 and multiple occurrences of the same existential variable are counted multiple times. Because this variant is not robust to slight changes in the goal, its use is strongly discouraged.
-
Variant
instantiate ( natural := term ) in ident -
Variant
instantiate ( natural := term ) in ( value of ident ) -
Variant
instantiate ( natural := term ) in ( type of ident ) These allow to refer respectively to existential variables occurring in a hypothesis or in the body or the type of a local definition (named
ident).
-
Variant
instantiate This tactic behaves functionally as
idtac.Deprecated since version 8.16.
-
Tactic
admit¶ This tactic allows temporarily skipping a subgoal so as to progress further in the rest of the proof. A proof containing admitted goals cannot be closed with
Qedbut only withAdmitted.
-
Variant
give_up Synonym of
admit.
-
Tactic
absurd term¶ This tactic applies to any goal. The argument term is any proposition
Pof typeProp. This tactic applies False elimination, that is it deduces the current goal from False, and generates as subgoals∼PandP. It is very useful in proofs by cases, where some cases are impossible. In most cases,Por∼Pis one of the hypotheses of the local context.
-
Tactic
contradiction¶ This tactic applies to any goal. The contradiction tactic attempts to find in the current context (after all intros) a hypothesis that is equivalent to an empty inductive type (e.g.
False), to the negation of a singleton inductive type (e.g.Trueorx=x), or two contradictory hypotheses.-
Error
No such assumption.¶
-
Error
-
Tactic
contradict ident¶ This tactic allows manipulating negated hypothesis and goals. The name
identshould correspond to a hypothesis. Withcontradict H, the current goal and context is transformed in the following way:H:¬A ⊢ B becomes ⊢ A
H:¬A ⊢ ¬B becomes H: B ⊢ A
H: A ⊢ B becomes ⊢ ¬A
H: A ⊢ ¬B becomes H: B ⊢ ¬A
-
Tactic
exfalso¶ This tactic implements the “ex falso quodlibet” logical principle: an elimination of False is performed on the current goal, and the user is then required to prove that False is indeed provable in the current context. This tactic is a macro for
elimtype False.
Classical tactics¶
In order to ease the proving process, when the Classical module is
loaded, a few more tactics are available. Make sure to load the module
using the Require Import command.
-
Tactic
classical_left¶ -
Tactic
classical_right¶ These tactics are the analog of
leftandrightbut using classical logic. They can only be used for disjunctions. Useclassical_leftto prove the left part of the disjunction with the assumption that the negation of right part holds. Useclassical_rightto prove the right part of the disjunction with the assumption that the negation of left part holds.
Performance-oriented tactic variants¶
-
Tactic
exact_no_check term¶ For advanced usage. Similar to
exactterm, but as an optimization, it skips checking thattermhas the goal's type, relying on the kernel check instead. Seechange_no_checkfor more explanation.Example
- Goal False.
- 1 goal ============================ False
- exact_no_check I.
- No more goals.
- Fail Qed.
- The command has indeed failed with message: The term "I" has type "True" while it is expected to have type "False".
-
Variant
vm_cast_no_check term¶ For advanced usage. Similar to
exact_no_checkterm, but additionally instructs the kernel to usevm_computeto compare the goal's type with theterm's type.Example
- Goal False.
- 1 goal ============================ False
- vm_cast_no_check I.
- No more goals.
- Fail Qed.
- The command has indeed failed with message: The term "I" has type "True" while it is expected to have type "False".
-
Variant
native_cast_no_check term¶ for advanced usage. similar to
exact_no_checkterm, but additionally instructs the kernel to usenative_computeto compare the goal's type with theterm's type.Example
- Goal False.
- 1 goal ============================ False
- native_cast_no_check I.
- No more goals.
- Fail Qed.
- The command has indeed failed with message: Native compiler is disabled, falling back to VM conversion test. [native-compiler-disabled,native-compiler]