Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improved sparse getindex #7131

Merged
merged 1 commit into from
Jun 5, 2014
Merged

Improved sparse getindex #7131

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
164 changes: 61 additions & 103 deletions base/sparse/sparsematrix.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -712,31 +712,40 @@ prod{T}(A::SparseMatrixCSC{T}, region) = reducedim(*,A,region,one(T))
#sum(A::SparseMatrixCSC{Bool}) = countnz(A)

## getindex
getindex(A::SparseMatrixCSC, i::Integer) = getindex(A, ind2sub(size(A),i)...)

function getindex{T}(A::SparseMatrixCSC{T}, i0::Integer, i1::Integer)
if !(1 <= i0 <= A.m && 1 <= i1 <= A.n); throw(BoundsError()); end
first = A.colptr[i1]
last = A.colptr[i1+1]-1
while first <= last
mid = (first + last) >> 1
t = A.rowval[mid]
if t == i0
return A.nzval[mid]
elseif t > i0
last = mid - 1
function binarysearch(haystack::AbstractVector, needle, lo::Int, hi::Int)
# Finds the first occurrence of needle in haystack[lo:hi]
lo = lo-1
hi2 = hi
hi = hi+1
@inbounds while lo < hi-1
m = (lo+hi)>>>1
if haystack[m] < needle
lo = m
else
first = mid + 1
hi = m
end
end
return zero(T)
(hi==hi2+1 || haystack[hi]!=needle) ? -1 : hi
end
function rangesearch(haystack::Range, needle)
(i,rem) = divrem(needle - first(haystack), step(haystack))
(rem==0 && 1<=i+1<=length(haystack)) ? i+1 : -1
end

getindex(A::SparseMatrixCSC, i::Integer) = getindex(A, ind2sub(size(A),i))
getindex(A::SparseMatrixCSC, I::(Integer,Integer)) = getindex(A, I[1], I[2])

function getindex{T}(A::SparseMatrixCSC{T}, i0::Integer, i1::Integer)
if !(1 <= i0 <= A.m && 1 <= i1 <= A.n); throw(BoundsError()); end
ind = binarysearch(A.rowval, i0, A.colptr[i1], A.colptr[i1+1]-1)
ind > -1 ? A.nzval[ind] : zero(T)
end

getindex{T<:Integer}(A::SparseMatrixCSC, I::AbstractVector{T}, j::Integer) = getindex(A,I,[j])
getindex{T<:Integer}(A::SparseMatrixCSC, i::Integer, J::AbstractVector{T}) = getindex(A,[i],J)

function getindex_cols{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, J::AbstractVector)

# for indexing whole columns
(m, n) = size(A)
nJ = length(J)

Expand All @@ -758,52 +767,35 @@ function getindex_cols{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, J::AbstractVector)

for j = 1:nJ
col = J[j]

for k = colptrA[col]:colptrA[col+1]-1
ptrS += 1
rowvalS[ptrS] = rowvalA[k]
nzvalS[ptrS] = nzvalA[k]
end
end

return SparseMatrixCSC(m, nJ, colptrS, rowvalS, nzvalS)

end

# TODO: See if growing arrays is faster than pre-computing structure
# and then populating nonzeros
# TODO: Use binary search in cases where nI >> nnz(A[:,j]) or nI << nnz(A[:,j])
function getindex_I_sorted{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::AbstractVector, J::AbstractVector)

function getindex{Tv,Ti<:Integer}(A::SparseMatrixCSC{Tv,Ti}, I::Range, J::AbstractVector)
# Ranges for indexing rows
(m, n) = size(A)
# whole columns:
if I == 1:m
return getindex_cols(A, J)
end

nI = length(I)
nJ = length(J)

colptrA = A.colptr; rowvalA = A.rowval; nzvalA = A.nzval

I_ref = falses(m)
I_ref[I] = true

I_repeat = zeros(Int, m)
for i=1:nI; I_repeat[I[i]] += 1; end

colptrS = Array(Ti, nJ+1)
colptrS[1] = 1
nnzS = 0

# Form the structure of the result and compute space
for j = 1:nJ
col = J[j]

for k = colptrA[col]:colptrA[col+1]-1
rowA = rowvalA[k]

if I_ref[rowA]
for r = 1:I_repeat[rowA]
nnzS += 1
end
end

for k in colptrA[col]:colptrA[col+1]-1
if rowvalA[k] in I; nnzS += 1 end # `in` is fast for ranges
end
colptrS[j+1] = nnzS+1
end
Expand All @@ -813,54 +805,40 @@ function getindex_I_sorted{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::AbstractVector,
nzvalS = Array(Tv, nnzS)
ptrS = 1

fI = zeros(Ti, m)
for k=1:nI
Ik = I[k]
if fI[Ik] == 0; fI[Ik] = k; end
end

for j = 1:nJ
col = J[j]

for k = colptrA[col]:colptrA[col+1]-1
rowA = rowvalA[k]

if I_ref[rowA]
for r = 1:I_repeat[rowA]
rowvalS[ptrS] = fI[rowA] + r - 1
nzvalS[ptrS] = nzvalA[k]
ptrS += 1
end
i = rangesearch(I, rowA)
if i > -1
rowvalS[ptrS] = i
nzvalS[ptrS] = nzvalA[k]
ptrS += 1
end

end
end

return SparseMatrixCSC(nI, nJ, colptrS, rowvalS, nzvalS)
end

# getindex_I_sorted based on merging of sorted lists
function getindex_I_sorted_old{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::Vector, J::AbstractVector)

function getindex_I_sorted{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::AbstractVector, J::AbstractVector)
# Sorted vectors for indexing rows.
# Similar to getindex_general but without the transpose trick.
(m, n) = size(A)
nI = length(I)
nJ = length(J)

colptrA = A.colptr; rowvalA = A.rowval; nzvalA = A.nzval

colptrS = Array(Ti, nJ+1)
colptrS[1] = 1
nnzS = 0

# Form the structure of the result and compute space
for j = 1:nJ
col = J[j]

ptrI::Int = 1

ptrI::Int = 1 # runs through I
ptrA::Int = colptrA[col]
stopA::Int = colptrA[col+1]

while ptrI <= nI && ptrA < stopA
rowA = rowvalA[ptrA]
rowI = I[ptrI]
Expand All @@ -874,22 +852,17 @@ function getindex_I_sorted_old{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::Vector, J::A
ptrI += 1
end
end
colptrS[j+1] = nnzS+1

colptrS[j+1] = nnzS+1
end

fI = find(I)

# Populate the values in the result
rowvalS = Array(Ti, nnzS)
nzvalS = Array(Tv, nnzS)
ptrS = 0
ptrS = 1

for j = 1:nJ
col = J[j]

ptrI::Int = 1

ptrI::Int = 1 # runs through I
ptrA::Int = colptrA[col]
stopA::Int = colptrA[col+1]

Expand All @@ -902,30 +875,27 @@ function getindex_I_sorted_old{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::Vector, J::A
elseif rowI < rowA
ptrI += 1
else
ptrS += 1
rowvalS[ptrS] = fI[ptrI]
rowvalS[ptrS] = ptrI
nzvalS[ptrS] = nzvalA[ptrA]
ptrS += 1
ptrI += 1
end
end

end

return SparseMatrixCSC(nI, nJ, colptrS, rowvalS, nzvalS)
end

function getindex_general{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::Vector, J::AbstractVector)
function getindex_general{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::AbstractVector, J::AbstractVector)
# Anything for indexing rows.
# This sorts I first then does some trick with constructing the transpose.
(m, n) = size(A)
nI = length(I)
nJ = length(J)

colptrA = A.colptr; rowvalA = A.rowval; nzvalA = A.nzval

nnzS = 0

pI = sortperm(I); I = I[pI]
fI = find(I)

W = zeros(Int, nI + 1) # Keep row counts
W[1] = 1 # For cumsum later

Expand Down Expand Up @@ -993,36 +963,24 @@ function getindex_general{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::Vector, J::Abstra
# Transpose so that rows are in sorted order and return
S_T = SparseMatrixCSC(nJ, nI, colptrS_T, rowvalS_T, nzvalS_T)
return S_T.'

end

# S = A[I, J]

# the general case:
function getindex{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, I::AbstractVector, J::AbstractVector)
m = size(A, 1)
if isa(I, Range)
if I == 1:m # whole columns
return getindex_cols(A, J)
else # ranges are always sorted, but maybe in reverse
if step(I)>0
return getindex_I_sorted(A, I, J)
else
I = [I]
return getindex_general(A, I, J)
# todo:
# return reverse(getindex_I_sorted(A, reverse(I), J))
end
end
else
if issorted(I)
return getindex_I_sorted(A, I, J)
else
return getindex_general(A, I, J)
end
loop_overI_threshold = 0 # ~40 but doesn't matter much.
if issorted(I)
return getindex_I_sorted(A, I, J)
else
return getindex_general(A, I, J)
end
end

# logical getindex
getindex{Tv,Ti<:Integer}(A::SparseMatrixCSC{Tv,Ti}, I::Range{Bool}, J::AbstractVector{Bool}) = error("Cannot index with Range{Bool}")
getindex{Tv,Ti<:Integer,T<:Integer}(A::SparseMatrixCSC{Tv,Ti}, I::Range{Bool}, J::AbstractVector{T}) = error("Cannot index with Range{Bool}")

getindex{T<:Integer}(A::SparseMatrixCSC, I::Range{T}, J::AbstractVector{Bool}) = A[I,find(J)]
getindex(A::SparseMatrixCSC, I::Integer, J::AbstractVector{Bool}) = A[I,find(J)]
getindex(A::SparseMatrixCSC, I::AbstractVector{Bool}, J::Integer) = A[find(I),J]
getindex(A::SparseMatrixCSC, I::AbstractVector{Bool}, J::AbstractVector{Bool}) = A[find(I),find(J)]
Expand Down
51 changes: 51 additions & 0 deletions test/sparse.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -239,6 +239,57 @@ for op in (:sin, :cos, :tan, :iceil, :ifloor, :ceil, :floor, :abs, :abs2)
end
end

# getindex tests
ni = 23
nj = 32
a116 = reshape(1:(ni*nj), ni, nj)
s116 = sparse(a116)

ad116 = diagm(diag(a116))
sd116 = sparse(ad116)

for (aa116, ss116) in [(a116, s116), (ad116, sd116)]
ij=11; i=3; j=2
@test ss116[ij] == aa116[ij]
@test ss116[(i,j)] == aa116[i,j]
@test ss116[i,j] == aa116[i,j]
@test ss116[i-1,j] == aa116[i-1,j]
ss116[i,j] = 0
@test ss116[i,j] == 0
ss116 = sparse(aa116)

# range indexing
@test full(ss116[i,:]) == aa116[i,:]
@test full(ss116[:,j]) == aa116[:,j]'' # sparse matrices/vectors always have ndims==2:
@test full(ss116[i,1:2:end]) == aa116[i,1:2:end]
@test full(ss116[1:2:end,j]) == aa116[1:2:end,j]''
@test full(ss116[i,end:-2:1]) == aa116[i,end:-2:1]
@test full(ss116[end:-2:1,j]) == aa116[end:-2:1,j]''
# float-range indexing is not supported

# sorted vector indexing
@test full(ss116[i,[3:2:end-3]]) == aa116[i,[3:2:end-3]]
@test full(ss116[[3:2:end-3],j]) == aa116[[3:2:end-3],j]''
@test full(ss116[i,[end-3:-2:1]]) == aa116[i,[end-3:-2:1]]
@test full(ss116[[end-3:-2:1],j]) == aa116[[end-3:-2:1],j]''

# unsorted vector indexing with repetition
p = [4, 1, 2, 3, 2, 6]
@test full(ss116[p,:]) == aa116[p,:]
@test full(ss116[:,p]) == aa116[:,p]
@test full(ss116[p,p]) == aa116[p,p]

# bool indexing
li = randbool(size(aa116,1))
lj = randbool(size(aa116,2))
@test full(ss116[li,j]) == aa116[li,j]''
@test full(ss116[li,:]) == aa116[li,:]
@test full(ss116[i,lj]) == aa116[i,lj]
@test full(ss116[:,lj]) == aa116[:,lj]
@test full(ss116[li,lj]) == aa116[li,lj]
end


# setindex tests
let a = spzeros(Int, 10, 10)
@test countnz(a) == 0
Expand Down