obtain

Summary

The obtain tactic can be used to do have and cases all in one go.

Example

If you have a hypothesis h : ∀ ε > 0, ∃ (N : ℕ), (1 : ℝ) / (N + 1) < ε, then you could specialize h to the case ε = 0.01 with have h2 := h 0.01 (by norm_num) (or specialize h 0.01 (by norm_num) if you’re prepared to change h) and then you can get to N and the hypothesis hN : 1 / (N + 1) < ε with cases h2 with N hN or rcases h2 with ⟨N, hN⟩. But you can do both steps in one go with obtain ⟨N, hN⟩ := h 0.01 (by norm_num).

To make your code more readable you can explicitly mention the type of ⟨N, hN⟩. In the above example you could write obtain ⟨N, hN⟩ : ∃ (N : ℕ), (1 : ℝ) / (N + 1) < 0.01 := h 0.01 (by norm_num), meaning that the reader can instantly see what the type of hN is.

Details

Note that, like rcases and rintro, obtain works up to definitional equality.

As with rcases and rintro, there is a rfl trick, where you can eliminate a variable using rfl instead of a hypothesis name.