The type of rules that specify how metadata for projections in changes. See initialize_simps_projection.

The @[_simps_str] attribute specifies the preferred projections of the given structure, used by the @[simps] attribute.

• This will usually be tagged by the @[simps] tactic.
• You can also generate this with the command initialize_simps_projections.
• To change the default value, see Note [custom simps projection].
• You are strongly discouraged to add this attribute manually.
• The first argument is the list of names of the universe variables used in the structure
• The second argument is a list that consists of the projection data for each projection.

The @[notation_class] attribute specifies that this is a notation class, and this notation should be used instead of projections by @[simps].

• The first argument tt for notation classes and ff for classes applied to the structure, like has_coe_to_sort and has_coe_to_fun
• The second argument is the name of the projection (by default it is the first projection of the structure)

Returns the projection information of a structure.

Auxiliary function of get_composite_of_projections.

Given a structure str and a projection proj, that could be multiple nested projections (separated by _), returns an expression that is the composition of these projections and a list of natural numbers, that are the projection numbers of the applied projections.

Get the projections used by simps associated to a given structure str.

The returned information is also stored in a parameter of the attribute @[_simps_str], which is given to str. If str already has this attribute, the information is read from this attribute instead. See the documentation for this attribute for the data this tactic returns.

The returned universe levels are the universe levels of the structure. For the projections there are three cases

• If the declaration {structure_name}.simps.{projection_name} has been declared, then the value of this declaration is used (after checking that it is definitionally equal to the actual projection. If you rename the projection name, the declaration should have the new projection name.
• You can also declare a custom projection that is a composite of multiple projections.
• Otherwise, for every class with the notation_class attribute, and the structure has an instance of that notation class, then the projection of that notation class is used for the projection that is definitionally equal to it (if there is such a projection). This means in practice that coercions to function types and sorts will be used instead of a projection, if this coercion is definitionally equal to a projection. Furthermore, for notation classes like has_mul and has_zero those projections are used instead of the corresponding projection. Projections for coercions and notation classes are not automatically generated if they are composites of multiple projections (for example when you use extend without the old_structure_cmd).
• Otherwise, the projection of the structure is chosen. For example: simps_get_raw_projections env `prod gives the default projections

  ([u, v], [prod.fst.{u v}, prod.snd.{u v}])

while ``simps_get_raw_projections env `equiv`` gives

  ([u_1, u_2], [λ α β, coe_fn, λ {α β} (e : α ≃ β), ⇑(e.symm), left_inv, right_inv])

after declaring the coercion from `equiv` to function and adding the declaration

  def equiv.simps.inv_fun {α β} (e : α ≃ β) : β → α := e.symm

Optionally, this command accepts three optional arguments:

• If trace_if_exists the command will always generate a trace message when the structure already has the attribute @[_simps_str].
• The rules argument accepts a list of pairs sum.inl (old_name, new_name). This is used to change the projection name old_name to the custom projection name new_name. Example: for the structure equiv the projection to_fun could be renamed apply. This name will be used for parsing and generating projection names. This argument is ignored if the structure already has an existing attribute. If an element of rules is of the form sum.inr name, this means that the projection name will not be applied by default.
• if trc is true, this tactic will trace information.

Parse a rule for initialize_simps_projections. It is either <name>→<name> or -<name>, possibly following by as_prefix.

Specify simps projections, see Note [custom simps projection].

• You can specify custom names by writing e.g. initialize_simps_projections equiv (to_fun → apply, inv_fun → symm_apply).
• You can disable a projection by default by running initialize_simps_projections equiv (-inv_fun) This will ensure that no simp lemmas are generated for this projection, unless this projection is explicitly specified by the user.
• If you want the projection name added as a prefix in the generated lemma name, you can add the as_prefix modifier: initialize_simps_projections equiv (to_fun → coe as_prefix) Note that this does not influence the parsing of projection names: if you have a declaration foo and you want to apply the projections snd, coe (which is a prefix) and fst, in that order you can run @[simps snd_coe_fst] def foo ... and this will generate a lemma with the name coe_foo_snd_fst.
  • Run initialize_simps_projections? (or set trace.simps.verbose) to see the generated projections.

A common configuration for @[simps]: generate equalities between functions instead equalities between fully applied expressions.

A common configuration for @[simps]: don't tag the generated lemmas with @[simp].

Get the projections of a structure used by @[simps] applied to the appropriate arguments. Returns a list of tuples

(corresponding right-hand-side, given projection name, projection expression, projection numbers,
  used by default, is prefix)

(where all fields except the first are packed in a projection_data structure) one for each projection. The given projection name is the name for the projection used by the user used to generate (and parse) projection names. For example, in the structure

Example 1: simps_get_projection_exprs env `(α × β) `(⟨x, y⟩) will give the output

  [(`(x), `fst, `(@prod.fst.{u v} α β), [0], tt, ff),
   (`(y), `snd, `(@prod.snd.{u v} α β), [1], tt, ff)]

Example 2: simps_get_projection_exprs env `(α ≃ α) `(⟨id, id, λ _, rfl, λ _, rfl⟩) will give the output

  [(`(id), `apply, `(coe), [0], tt, ff),
   (`(id), `symm_apply, `(λ f, ⇑f.symm), [1], tt, ff),
   ...,
   ...]

Add a lemma with nm stating that lhs = rhs. type is the type of both lhs and rhs, args is the list of local constants occurring, and univs is the list of universe variables. If add_simp then we make the resulting lemma a simp lemma.

Derive lemmas specifying the projections of the declaration. If todo is non-empty, it will generate exactly the names in todo. to_apply is non-empty after a custom projection that is a composition of multiple projections was just used. In that case we need to apply these projections before we continue changing lhs.

simps_tac derives simp lemmas for all (nested) non-Prop projections of the declaration. If todo is non-empty, it will generate exactly the names in todo. If short_nm is true, the generated names will only use the last projection name. If trc is true, trace as if trace.simps.verbose is true.

The parser for the @[simps] attribute.

The @[simps] attribute automatically derives lemmas specifying the projections of this declaration.

Example:

@[simps] def foo : ℕ × ℤ := (1, 2)

derives two simp lemmas:

@[simp] lemma foo_fst : foo.fst = 1
@[simp] lemma foo_snd : foo.snd = 2

• It does not derive simp lemmas for the prop-valued projections.

• It will automatically reduce newly created beta-redexes, but will not unfold any definitions.

• If the structure has a coercion to either sorts or functions, and this is defined to be one of the projections, then this coercion will be used instead of the projection.

• If the structure is a class that has an instance to a notation class, like has_mul, then this notation is used instead of the corresponding projection.

• You can specify custom projections, by giving a declaration with name {structure_name}.simps.{projection_name}. See Note [custom simps projection].

  Example:

  def equiv.simps.inv_fun (e : α ≃ β) : β → α := e.symm
  @[simps] def equiv.trans (e₁ : α ≃ β) (e₂ : β ≃ γ) : α ≃ γ :=
  ⟨e₂ ∘ e₁, e₁.symm ∘ e₂.symm⟩
  

  generates

  @[simp] lemma equiv.trans_to_fun : ∀ {α β γ} (e₁ e₂) (a : α), ⇑(e₁.trans e₂) a = (⇑e₂ ∘ ⇑e₁) a
  @[simp] lemma equiv.trans_inv_fun : ∀ {α β γ} (e₁ e₂) (a : γ),
    ⇑((e₁.trans e₂).symm) a = (⇑(e₁.symm) ∘ ⇑(e₂.symm)) a
  

• You can specify custom projection names, by specifying the new projection names using initialize_simps_projections. Example: initialize_simps_projections equiv (to_fun → apply, inv_fun → symm_apply).

• If one of the fields itself is a structure, this command will recursively create simp lemmas for all fields in that structure.

  • Exception: by default it will not recursively create simp lemmas for fields in the structures prod and pprod. Give explicit projection names to override this behavior.

  Example:

  structure my_prod (α β : Type*) := (fst : α) (snd : β)
  @[simps] def foo : prod ℕ ℕ × my_prod ℕ ℕ := ⟨⟨1, 2⟩, 3, 4⟩
  

  generates

  @[simp] lemma foo_fst : foo.fst = (1, 2)
  @[simp] lemma foo_snd_fst : foo.snd.fst = 3
  @[simp] lemma foo_snd_snd : foo.snd.snd = 4
  

• You can use @[simps proj1 proj2 ...] to only generate the projection lemmas for the specified projections.

• Recursive projection names can be specified using proj1_proj2_proj3. This will create a lemma of the form foo.proj1.proj2.proj3 = ....

  Example:

  structure my_prod (α β : Type*) := (fst : α) (snd : β)
  @[simps fst fst_fst snd] def foo : prod ℕ ℕ × my_prod ℕ ℕ := ⟨⟨1, 2⟩, 3, 4⟩
  

  generates

  @[simp] lemma foo_fst : foo.fst = (1, 2)
  @[simp] lemma foo_fst_fst : foo.fst.fst = 1
  @[simp] lemma foo_snd : foo.snd = {fst := 3, snd := 4}
  

• If one of the values is an eta-expanded structure, we will eta-reduce this structure.

  Example:

  structure equiv_plus_data (α β) extends α ≃ β := (data : bool)
  @[simps] def bar {α} : equiv_plus_data α α := { data := tt, ..equiv.refl α }
  

  generates the following, even though Lean inserts an eta-expanded version of equiv.refl α in the definition of bar:

  @[simp] lemma bar_to_equiv : ∀ {α : Sort u_1}, bar.to_equiv = equiv.refl α
  @[simp] lemma bar_data : ∀ {α : Sort u_1}, bar.data = tt
  

• For configuration options, see the doc string of simps_cfg.

• The precise syntax is ('simps' ident* e), where e is an expression of type simps_cfg.

• @[simps] reduces let-expressions where necessary.

• If one of the fields is a partially applied constructor, we will eta-expand it (this likely never happens).

• When option trace.simps.verbose is true, simps will print the projections it finds and the lemmas it generates. The same can be achieved by using @[simps?], except that in this case it will not print projection information.

• Use @[to_additive, simps] to apply both to_additive and simps to a definition, making sure that simps comes after to_additive. This will also generate the additive versions of all simp lemmas. Note however, that the additive versions of the simp lemmas always use the default name generated by to_additive, even if a custom name is given for the additive version of the definition.