Stream: Isabelle/ML

Topic: combine premises in goal


view this post on Zulip Jan van Brügge (Jan 22 2022 at 11:38):

Given a goal like this:

⋀a x. f x ⟹ x = T a ⟹ Q

what is the best way to rewrite the premises to

⋀a. f (T a) ⟹ Q

? I could hack something together with SUBPROOF but there has to be a better way

view this post on Zulip Lukas Stevens (Jan 22 2022 at 11:56):

You could do this:

lemma a: "(⋀a x. f x ⟹ x = T a ⟹ PROP Q) ≡ (⋀a. f (T a) ⟹ PROP Q)"
proof
  fix a assume "⋀a x. f x ⟹ x = T a ⟹ PROP Q" "f (T a)"
  from this(1)[OF this(2), of a] show "PROP Q" by simp
next
  fix a x assume "⋀a. f (T a) ⟹ PROP Q" "f x" "x = T a"
  from this(1)[of a] this(2,3) show "PROP Q" by simp
qed

lemma "⋀a x. f x ⟹ x = T a ⟹ Q"
  apply(tactic CONVERSION (Conv.rewr_conv @{thm a}) 1)

view this post on Zulip Jan van Brügge (Jan 22 2022 at 11:57):

I don't know f nor T statically, so this won't work. I guess the SUBPROOF hack it is

view this post on Zulip Lukas Stevens (Jan 22 2022 at 11:58):

But this works for any f and T?

view this post on Zulip Jan van Brügge (Jan 22 2022 at 11:58):

oh, yes of course, I am stupid

view this post on Zulip Jan van Brügge (Jan 22 2022 at 12:32):

Weird, now I am at ⋀b3 z3. b3 = z3 ⟹ b3 = z3 but assume_tac fails

view this post on Zulip Lukas Stevens (Jan 22 2022 at 12:34):

The types do match, right?

view this post on Zulip Jan van Brügge (Jan 22 2022 at 12:34):

yes

view this post on Zulip Jan van Brügge (Jan 22 2022 at 12:34):

all are 'c

view this post on Zulip Lukas Stevens (Jan 22 2022 at 12:47):

You can try printing the term literally without pretty-printing. Also the context you pass in is the right one?

view this post on Zulip Jan van Brügge (Jan 22 2022 at 12:52):

Const ("Pure.all", "('c ⇒ prop) ⇒ prop") $
  Abs ("b3", "'c",
    Const ("Pure.all", "('c ⇒ prop) ⇒ prop") $
      Abs ("z3", "'c",
        Const ("Pure.imp", "prop ⇒ prop ⇒ prop") $
          (Const ("HOL.Trueprop", "bool ⇒ prop") $ (Const ("HOL.eq", "'c ⇒ 'c ⇒ bool") $ Bound 1 $ Bound 0)) $
          (Const ("HOL.Trueprop", "bool ⇒ prop") $
            (Const ("HOL.eq", "'c ⇒ 'c ⇒ bool") $ Bound 1 $ Bound 0))))

view this post on Zulip Jan van Brügge (Jan 22 2022 at 12:52):

Yes, there is only one context. Also apply assumption after the tactic solves the goal as expected


Last updated: Jul 15 2022 at 23:21 UTC