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	0.735717	0.839505	0.0154341	0.675586	0.881035
2	0.42152	0.680851	0.395228	0.727355	0.828121
3	0.650958	0.800487	0.252657	0.585549	0.112034

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	0.662533	0.0791648	0.476085	0.137527	0.780112	0.556826	0.29654	0.0462489	0.580725	0.459269
2	0.641919	0.219975	0.262794	0.391827	0.116199	0.397834	0.894128	0.178664	0.84901	0.97465
3	0.633747	0.404336	0.422398	0.523208	0.839493	0.994638	0.938289	0.109516	0.35902	0.153182
4	0.878339	0.600868	0.341619	0.536543	0.362645	0.796499	0.88459	0.849896	0.293786	0.657048
5	0.106658	0.150221	0.91712	0.809954	0.617408	0.316943	0.958499	0.409879	0.489289	0.719617
6	0.59303	0.275082	0.659807	0.987099	0.919333	0.0796179	0.496563	0.714089	0.260569	0.889253
7	0.556824	0.114174	0.876116	0.833703	0.583783	0.358228	0.47351	0.891788	0.932091	0.851929
8	0.324652	0.69072	0.086412	0.690169	0.713064	0.22064	0.597105	0.319834	0.191315	0.122917
9	0.397104	0.963838	0.336063	0.429932	0.799338	0.543028	0.716992	0.427673	0.572117	0.849775
10	0.830195	0.716072	0.704308	0.363882	0.0986513	0.659677	0.596932	0.998389	0.236294	0.615016
11	0.328347	0.667366	0.0899753	0.868009	0.579315	0.633655	0.931564	0.815405	0.526418	0.981884
12	0.881655	0.453405	0.154194	0.566519	0.352452	0.985721	0.723605	0.827787	0.880202	0.767686
13	0.636864	0.673021	0.249024	0.710393	0.475797	0.128731	0.359529	0.315217	0.723981	0.812359
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮
999989	0.975986	0.476689	0.0514201	0.989957	0.883476	0.357384	0.243844	0.916654	0.496581	0.0234581
999990	0.0708377	0.110511	0.594757	0.679725	0.334746	0.211194	0.680969	0.165843	0.257112	0.459214
999991	0.579021	0.539474	0.200064	0.594268	0.462375	0.27947	0.229069	0.024284	0.37251	0.296529
999992	0.191944	0.0983965	0.252409	0.793691	0.375082	0.973486	0.845244	0.0734637	0.777887	0.722518
999993	0.542728	0.92741	0.0118351	0.110601	0.527383	0.96193	0.958437	0.634349	0.523464	0.171206
999994	0.423901	0.768264	0.813564	0.647029	0.776164	0.744056	0.304229	0.2424	0.905494	0.639531
999995	0.641797	0.871426	0.213085	0.805663	0.341525	0.524431	0.393226	0.0950247	0.158032	0.357649
999996	0.153543	0.311477	0.508389	0.424192	0.512524	0.0338346	0.620033	0.749842	0.29848	0.116878
999997	0.494436	0.126412	0.295339	0.472308	0.110807	0.639409	0.10246	0.0628602	0.502538	0.0265686
999998	0.418533	0.756071	0.936742	0.142512	0.231282	0.509376	0.686532	0.602506	0.361023	0.378734
999999	0.0709701	0.838043	0.807257	0.0562733	0.775084	0.251166	0.895086	0.38131	0.652535	0.614597
1000000	0.259408	0.747359	0.616164	0.749795	0.622107	0.0136075	0.959057	0.192052	0.958715	0.629737

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

  2.248203 seconds (60.12 M allocations: 1.056 GiB, 17.66% gc time, 4.00% compilation time)

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

  1.672558 seconds (59.99 M allocations: 1.050 GiB, 5.47% gc time)

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

  0.271730 seconds (1.22 M allocations: 66.497 MiB, 4.36% gc time, 93.31% compilation time)

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

  0.007500 seconds (21 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

  0.434296 seconds (712.75 k allocations: 43.180 MiB, 97.27% compilation time: 94% of which was recompilation)

1000000-element Vector{Float64}:
 4.075031145560174
 4.9269995567112765
 5.37782807808806
 6.201831198796953
 5.49558611192314
 5.874443898489834
 6.472146151299726
 3.956827075489279
 6.035858171073009
 5.819416611082717
 ⋮
 5.104121438607423
 5.369343756593169
 6.2646325534129454
 4.4018586929999035
 3.729193960552614
 2.833138329204522
 5.023312467271604
 5.3423217101664395
 5.748001588947343

Do it again

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

  0.011397 seconds (123 allocations: 7.634 MiB)

1000000-element Vector{Float64}:
 4.075031145560174
 4.9269995567112765
 5.37782807808806
 6.201831198796953
 5.49558611192314
 5.874443898489834
 6.472146151299726
 3.956827075489279
 6.035858171073009
 5.819416611082717
 ⋮
 5.104121438607423
 5.369343756593169
 6.2646325534129454
 4.4018586929999035
 3.729193960552614
 2.833138329204522
 5.023312467271604
 5.3423217101664395
 5.748001588947343

---

This notebook was generated using Literate.jl.