-
Notifications
You must be signed in to change notification settings - Fork 0
/
column.jl
174 lines (131 loc) · 4.89 KB
/
column.jl
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
"""
DFColumn{T}
Lazy representation of table column. Do not instantate it directly, use indexing of table or view.
# Notes
DFColumn is not AbstractVector, but it support iteration and getindex. Iteration is much more efficient then consequentially
get index.
You can materialize DFColumn to Vector with [materialize(c::DFColumn)](@ref)
Broadcasting of DFColumn also supported. Broadcast which arguments is DFColumns and, optionally, scalars is DFColumn too.
# Examples
```julia
t = open_table("test_table") #table with `price` column
col = t.price
count = length(col)
price_condition = 10 .< t.price .< 40
in_condition_count = sum(price_condition)
in_condition = col[price_condition]
first_100_in_condition_vector = materialize(in_condition[1:100])
```
"""
struct DFColumn{T}
view::DFView
function DFColumn(view::DFView)
size(view, 2) != 1 && throw(ArgumentError("Column projection must contains singe element"))
new{coltype(view.projection, 1)}(view)
end
end
Base.:(==)(a::DFColumn, b::DFColumn) = a.view == b.view
Base.show(io::IO, c::DFColumn) = print(io, typeof(c))
Base.eltype(::Type{DFColumn{T}}) where {T} = T
Base.eltype(::DFColumn{T}) where {T} = T
Base.size(c::DFColumn) = (nrow(c.view),)
Base.size(c::DFColumn, dim::Number) = dim == 1 ? nrow(c.view) : 1
Base.length(c::DFColumn) = Base.size(c, 1)
Base.lastindex(c::DFColumn) = nrow(c.view)
Base.ndims(c::DFColumn) = 1
Base.ndims(c::Type{DFColumn}) = 1
Base.IndexStyle(::Type{<:DFColumn}) = IndexLinear()
function Base.getindex(c::DFColumn, i::AbstractRange)
DFColumn(selection(c.view, i))
end
function Base.getindex(c::DFColumn, i::DFColumn{Bool})
c.view.selection != i.view.selection && throw(ArgumentError("cols must have same selections"))
DFColumn(selection(c.view, i))
end
map_to_column(f::Function, c::DFColumn) = map_to_column(f, c.view)
function selection(v::DFView, col::DFColumn{Bool})
v.selection != col.view.selection && throw(ArgumentError("col must have same selection as view"))
selection(v, col.view.projection.cols[1])
end
function Base.setproperty!(v::DFView, name::Symbol, value::DFColumn)
!issameselection(v, value.view) && throw(ArgumentError("Can't add column with another selection"))
return v.projection = add(v.projection, (;(name=>value.view.projection.cols[1],)...,))
end
function Base.copyto!(dest::AbstractVector, src::DFColumn)
offset = 1
for block in BlocksIterator(src.view)
view(dest, offset:(offset + length(block[1]) - 1)) .= block[1]
offset += length(block[1])
end
return dest
end
function Base.getindex(c::DFColumn, i::Number)
tmp_view = selection(c.view, i)
it = BlocksIterator(tmp_view)
res = iterate(it)
isnothing(res) && throw(BoundsError(c, i))
return res[1][1][1]
end
function Base.iterate(c::DFColumn{T}) where {T}
it = BlocksIterator(c.view)
block_res = iterate(it)
isnothing(block_res) && return nothing
block_data = block_res[1][1]
inblock_pos = 1
return (
block_data[inblock_pos], (it, block_data, inblock_pos)
)
end
function Base.iterate(c::DFColumn, state)
(it, block_data, inblock_pos) = state
inblock_pos += 1
if inblock_pos <= length(block_data)
return (block_data[inblock_pos], (it, block_data, inblock_pos))
end
block_res = iterate(it)
isnothing(block_res) && return nothing
block_data = block_res[1][1]
return (
block_data[1], (it, block_data, 1)
)
end
"""
DFView(cols::NamedTuple{Cols, <:Tuple{Vararg{<:DFColumn}}})
DFView(;kwargs...)
Make new view from DFColums. All columns must have same selection
# Examples
```julia
df = DataFrame((a=collect(1:100), b = collect(1:100)))
t = create_table("test", from = df)
v = DFView(col1 = t.a .+ t.b, col2 = t.a .- t.b, col3 = t.a .* t.a)
materialize(v)
```
"""
function DFView(cols::NamedTuple{Cols, <:Tuple{Vararg{<:DFColumn}}}) where {Cols}
table_selection = nothing
for col in cols
if (isnothing(table_selection))
table_selection = (col.view.table, col.view.selection)
else
((col.view.table, col.view.selection) != table_selection) && throw(ArgumentError("All columns must have same selection and table"))
end
end
projection = Projection(
(;zip(Cols, map(c->c.view.projection.cols[1], cols))...)
)
return DFView(
table_selection[1],
projection,
table_selection[2]
)
end
DFView(;kwargs...) = DFView((;kwargs...,))
"""
add_column(v::DFView, name::Symbol, column::DFColum)
Add column to view. Column must have same selection as view
"""
function add_column!(v::DFView, name::Symbol, column::DFColumn)
!issameselection(v, column.view) && throw(ArgumentError("col must have same selection as view"))
v.projection = add(v.projection, NamedTuple{(name,)}((column.view.projection.cols[1],)))
return v
end