Correct some typos (#335)

Signed-off-by: Alexander Seiler <[email protected]>

docs/src/iterators.md

@@ -19,7 +19,7 @@ end
 Rather than calling `my_solver!(x, A, b)`, you could also initialize the iterable yourself and perform the `for` loop.
 
 ## Example: avoiding unnecessary initialization
-The Jacobi method for `SparseMatrixCSC` has some overhead in intialization; not only do we need to allocate a temporary vector, we also have to search for indices of the diagonal (and check their values are nonzero). The current implementation initializes the iterable as:
+The Jacobi method for `SparseMatrixCSC` has some overhead in initialization; not only do we need to allocate a temporary vector, we also have to search for indices of the diagonal (and check their values are nonzero). The current implementation initializes the iterable as:
 
 ```julia
 jacobi_iterable(x, A::SparseMatrixCSC, b; maxiter::Int = 10) =

src/history.jl

@@ -144,7 +144,7 @@ push_custom_data!(ch::CompleteHistory, key::Symbol, data) = ch.data[key][ch.iter
     reserve!(ch, key, maxiter, size)
     reserve!(typ, ch, key, maxiter, size)
 
-Reserve space for per iteration data in `ch`. If size is provided, intead of a
+Reserve space for per iteration data in `ch`. If size is provided, instead of a
 vector it will reserve matrix of dimensions `(maxiter, size)`.
 
 # Arguments

src/qmr.jl

@@ -233,7 +233,7 @@ Solves the problem ``Ax = b`` with the Quasi-Minimal Residual (QMR) method.
 
 ## Keywords
 
-- `initally_zero::Bool`: If `true` assumes that `iszero(x)` so that one
+- `initially_zero::Bool`: If `true` assumes that `iszero(x)` so that one
   matrix-vector product can be saved when computing the initial residual
   vector;
 - `maxiter::Int = size(A, 2)`: maximum number of iterations;

src/svdl.jl

@@ -516,7 +516,7 @@ full reorthogonalization. As explained in the numerical analysis literature by
 Kahan, Golub, Rutishauser, and others in the 1970s, double classical
 Gram-Schmidt reorthogonalization always suffices to keep vectors orthogonal to
 within machine precision. As described in [^Bjorck2015], `α` is a threshold
-for determinining when the second orthogonalization is necessary. -log10(α) is
+for determining when the second orthogonalization is necessary. -log10(α) is
 the number of (decimal) digits lost due to cancellation. Common choices are
 `α=0.1` [Rutishauser] and `α=1/√2` [Daniel1976] (our default).