Possible pitfalls

using DataFrames
using BenchmarkTools

Know what is copied when creating a DataFrame

x = DataFrame(rand(3, 5), :auto)

3×5 DataFrame

Row	x1	x2	x3	x4	x5
	Float64	Float64	Float64	Float64	Float64
1	0.947396	0.594914	0.318333	0.291359	0.591581
2	0.27621	0.856675	0.318142	0.795397	0.151377
3	0.78505	0.0548061	0.864647	0.708265	0.775993

x and y are not the same object

y = copy(x)
x === y

false

x and y are not the same object

y = DataFrame(x)
x === y

false

the columns are also not the same

any(x[!, i] === y[!, i] for i in ncol(x))

false

x and y are not the same object

y = DataFrame(x, copycols=false)
x === y

false

But the columns are the same

all(x[!, i] === y[!, i] for i in ncol(x))

true

the same when creating data frames using kwarg syntax

x = 1:3;
y = [1, 2, 3];
df = DataFrame(x=x, y=y)

3×2 DataFrame

Row	x	y
	Int64	Int64
1	1	1
2	2	2
3	3	3

different object

y === df.y

false

range is converted to a vector

typeof(x), typeof(df.x)

(UnitRange{Int64}, Vector{Int64})

slicing rows always creates a copy

y === df[:, :y]

false

you can avoid copying by using copycols=false keyword argument in functions.

df = DataFrame(x=x, y=y, copycols=false)

3×2 DataFrame

Row	x	y
	Int64	Int64
1	1	1
2	2	2
3	3	3

now it is the same

y === df.y

true

not the same object

select(df, :y)[!, 1] === y

false

the same object

select(df, :y, copycols=false)[!, 1] === y

true

Do not modify the parent of GroupedDataFrame or view

x = DataFrame(id=repeat([1, 2], outer=3), x=1:6)
g = groupby(x, :id)

x[1:3, 1] = [2, 2, 2]
g ## well - it is wrong now, g is only a view

GroupedDataFrame with 2 groups based on key: id

First Group (3 rows): id = 2

Row	id	x
	Int64	Int64
1	2	1
2	2	3
3	1	5

⋮

Last Group (3 rows): id = 2

Row	id	x
	Int64	Int64
1	2	2
2	2	4
3	2	6

s = view(x, 5:6, :)

2×2 SubDataFrame

Row	id	x
	Int64	Int64
1	1	5
2	2	6

delete!(x, 3:6)

2×2 DataFrame

Row	id	x
	Int64	Int64
1	2	1
2	2	2

This is an error

s ## Will return BoundsError

Single column selection for a DataFrame

Single column selection for a DataFrame creates aliases with ! and getproperty syntax and copies with :

x = DataFrame(a=1:3)
x.b = x[!, 1] ## alias
x.c = x[:, 1] ## copy
x.d = x[!, 1][:] ## copy
x.e = copy(x[!, 1]) ## explicit copy
display(x)

3×5 DataFrame

Row	a	b	c	d	e
	Int64	Int64	Int64	Int64	Int64
1	1	1	1	1	1
2	2	2	2	2	2
3	3	3	3	3	3

x[1, 1] = 100
display(x)

3×5 DataFrame

Row	a	b	c	d	e
	Int64	Int64	Int64	Int64	Int64
1	100	100	1	1	1
2	2	2	2	2	2
3	3	3	3	3	3

When iterating rows of a data frame

• use eachrow to avoid compilation cost in wide tables,

• but Tables.namedtupleiterator for fast execution in tall tables

The table below is tall:

df2 = DataFrame(rand(10^6, 10), :auto)

1000000×10 DataFrame

999975 rows omitted

Row	x1	x2	x3	x4	x5	x6	x7	x8	x9	x10
	Float64	Float64	Float64	Float64	Float64	Float64	Float64	Float64	Float64	Float64
1	0.538165	0.477619	0.63865	0.114941	0.591869	0.919326	0.460326	0.441607	0.0579152	0.00357446
2	0.533138	0.395225	0.936605	0.0296993	0.239957	0.31099	0.293299	0.950694	0.102088	0.362718
3	0.128026	0.037075	0.884437	0.150983	0.480751	0.819375	0.488905	0.0527264	0.08525	0.918672
4	0.0638011	0.676545	0.107159	0.344157	0.911609	0.852503	0.579901	0.129051	0.665324	0.439404
5	0.409903	0.117916	0.241907	0.676004	0.697972	0.169596	0.720508	0.653289	0.678077	0.266621
6	0.518344	0.960537	0.651609	0.67432	0.979009	0.873823	0.128961	0.0356952	0.139544	0.247548
7	0.717526	0.773986	0.564901	0.594871	0.60695	0.269118	0.809437	0.725244	0.383868	0.606193
8	0.437243	0.984836	0.500519	0.514212	0.130419	0.852367	0.511447	0.0222195	0.944033	0.438669
9	0.0227649	0.115036	0.0683505	0.0503512	0.943526	0.623187	0.0234799	0.461865	0.348735	0.690593
10	0.455792	0.527437	0.187127	0.679676	0.464027	0.414834	0.27163	0.717938	0.84506	0.394637
11	0.143232	0.592707	0.461296	0.0288561	0.433593	0.703271	0.286978	0.670572	0.570357	0.0609912
12	0.470649	0.324332	0.932704	0.548668	0.815837	0.877708	0.688376	0.445633	0.28391	0.827211
13	0.320294	0.38712	0.651804	0.278253	0.985029	0.777575	0.350404	0.403067	0.385916	0.495512
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮
999989	0.343766	0.349236	0.531518	0.591489	0.612046	0.0491963	0.328509	0.655394	0.605836	0.524536
999990	0.0673405	0.30125	0.954371	0.765194	0.646108	0.694336	0.358838	0.803799	0.0972333	0.119382
999991	0.519115	0.654394	0.84288	0.858733	0.541975	0.964638	0.921998	0.536797	0.101776	0.83639
999992	0.986049	0.888041	0.585754	0.817333	0.0598934	0.88705	0.916273	0.890539	0.639275	0.625951
999993	0.969635	0.271582	0.972013	0.988876	0.455172	0.888254	0.60208	0.431273	0.359801	0.607461
999994	0.617444	0.079036	0.67513	0.484139	0.00749603	0.383036	0.38363	0.0849654	0.510573	0.0714929
999995	0.710299	0.783795	0.977731	0.885664	0.731486	0.728469	0.406494	0.77531	0.520003	0.891723
999996	0.37826	0.579551	0.370413	0.286594	0.309514	0.685915	0.421191	0.224232	0.0486228	0.0447236
999997	0.0815508	0.767083	0.779161	0.728052	0.474848	0.475733	0.0454733	0.861135	0.162566	0.7554
999998	0.241736	0.215685	0.685861	0.400015	0.570262	0.0111432	0.69192	0.29754	0.521174	0.314071
999999	0.680218	0.933159	0.599354	0.116367	0.914086	0.136051	0.4714	0.171243	0.315983	0.879676
1000000	0.212074	0.266076	0.825468	0.923946	0.721343	0.081667	0.0335297	0.534697	0.53835	0.00548066

@time map(sum, eachrow(df2));

  2.705622 seconds (60.11 M allocations: 1.056 GiB, 25.15% gc time, 3.68% compilation time)

@time map(sum, eachrow(df2));

  1.862482 seconds (59.99 M allocations: 1.050 GiB, 5.09% gc time)

@time map(sum, Tables.namedtupleiterator(df2));

  0.348988 seconds (1.32 M allocations: 72.991 MiB, 2.33% gc time, 98.14% compilation time)

@time map(sum, Tables.namedtupleiterator(df2));

  0.006220 seconds (25 allocations: 7.631 MiB)

as you can see - this time it is much faster to iterate a type stable container still you might want to use the select syntax, which is optimized for such reductions:

this includes compilation time

@time select(df2, AsTable(:) => ByRow(sum) => "sum").sum

  0.632661 seconds (754.52 k allocations: 46.014 MiB, 99.11% compilation time: 96% of which was recompilation)

1000000-element Vector{Float64}:
 4.243993540615169
 4.154414077448179
 4.046201057064972
 4.769453432081732
 4.631793508413152
 5.2093905747165135
 6.052092977953647
 5.335964653974474
 3.3478892521388897
 4.958159796081296
 ⋮
 7.296159776774205
 6.546147719145592
 3.296942512182397
 7.410973094568378
 3.3490167652546234
 5.1310010949088385
 3.949406516653557
 5.217535862311809
 4.142632091909143

Do it again

@time select(df2, AsTable(:) => ByRow(sum) => "sum").sum

  0.004660 seconds (127 allocations: 7.634 MiB)

1000000-element Vector{Float64}:
 4.243993540615169
 4.154414077448179
 4.046201057064972
 4.769453432081732
 4.631793508413152
 5.2093905747165135
 6.052092977953647
 5.335964653974474
 3.3478892521388897
 4.958159796081296
 ⋮
 7.296159776774205
 6.546147719145592
 3.296942512182397
 7.410973094568378
 3.3490167652546234
 5.1310010949088385
 3.949406516653557
 5.217535862311809
 4.142632091909143

---

This notebook was generated using Literate.jl.