-
Notifications
You must be signed in to change notification settings - Fork 0
/
posit-conf-2023-recap.qmd
1129 lines (821 loc) · 30.9 KB
/
posit-conf-2023-recap.qmd
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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
---
title: "posit::conf(<FONT COLOR='white'>2023</FONT>)"
author: "Javier Orraca-Deatcu"
format:
revealjs:
embed-resources: false
theme: [default, styling.scss]
title-slide-attributes:
data-background-size: 360px
data-background-position: bottom 50px right 50px
data-background-color: "#76AADB"
eval: false
revealjs-plugins:
- codewindow
knitr: true
---
```{r}
#| label: setup
#| eval: true
#| include: false
#| file: setup.R
```
## Posit Conference
Posit, formerly RStudio, hosts an [annual conference](https://posit.co/conference/) including workshops, lunch-and-learns, and presentations touching on all things data science.
. . .
- 2023 conference was hosted in Chicago
- Approximately [1,200]{.fragment .highlight-teal} in-person attendees
. . .
R represented the majority of the presentations but there was more Python content than in prior years
::: footer
Posit | Learn more at [https://posit.co/](https://posit.co/)
:::
---
::: r-fit-text
Conference
:::
::: r-fit-text
Highlights
:::
# 1. Quarto
::: footer
Just to clarify... That's **Quarto**... **Q-U-A-R-T-O**, not #4 in Spanish (or "cuatro").
:::
## Quarto Presentations `r hexes_svg("quarto")`
These web slides were built with [Quarto](https://quarto.org/) and this section was copied almost entirely from Quarto's Reveal.js demo documentation
. . .
The next few slides cover what you can do with Quarto and [Reveal.js](https://revealjs.com) including:
- Presenting code and LaTeX equations
- Rendering code chunk computations in slide output
- Fancy transitions, animations, and code windows
::: footer
Quarto | Learn more on Quarto's Reveal.js [demo presentation](https://github.com/quarto-dev/quarto-web/blob/main/docs/presentations/revealjs/demo/index.qmd)
:::
## Pretty Code {auto-animate="true"}
- Over 20 syntax highlighting themes available
- Default theme optimized for accessibility
``` r
# Define a server for the Shiny app
function(input, output) {
# Fill in the spot we created for a plot
output$phonePlot <- renderPlot({
# Render a barplot
})
}
```
::: footer
Quarto | [Syntax Highlighting](https://quarto.org/docs/output-formats/html-code.html#highlighting)
:::
## Code Animations {auto-animate="true"}
- Over 20 syntax highlighting themes available
- Default theme optimized for accessibility
``` r
# Define a server for the Shiny app
function(input, output) {
# Fill in the spot we created for a plot
output$phonePlot <- renderPlot({
# Render a barplot
barplot(WorldPhones[,input$region]*1000,
main=input$region,
ylab="Number of Telephones",
xlab="Year")
})
}
```
::: footer
Quarto | [Code Animations](https://quarto.org/docs/presentations/revealjs/advanced.html#code-animations)
:::
## Line Highlighting
- Highlight specific lines for emphasis
- Incrementally highlight additional lines
``` {.python code-line-numbers="4-5|7|10"}
import numpy as np
import matplotlib.pyplot as plt
r = np.arange(0, 2, 0.01)
theta = 2 * np.pi * r
fig, ax = plt.subplots(subplot_kw={'projection': 'polar'})
ax.plot(theta, r)
ax.set_rticks([0.5, 1, 1.5, 2])
ax.grid(True)
plt.show()
```
::: footer
Quarto | [Line Highlighting](https://quarto.org/docs/presentations/revealjs/#line-highlighting)
:::
## Code Windows
::: columns
::: {.column width="50%"}
::: {.codewindow .quarto}
index.qmd
{{< include codewindow_files/example_include.qmd >}}
:::
:::
::: {.column width="50%"}
::: {.codewindow .html}
index.html
<iframe class="slide-deck" src="./codewindow_files/example.html" style="width: 100%; height: 484.47px;"></iframe>
:::
:::
:::
::: footer
Quarto | [codewindow](https://github.com/EmilHvitfeldt/quarto-revealjs-codewindow) Reveal.js extension & example code above from Emil Hvitfeldt's [Quarto Theming Talk at Posit Conf](https://github.com/EmilHvitfeldt/talk-quarto-theming-positconf)
:::
## Counters
Great for training purposes or when you've got a time limit
::: columns
::: {.column width="50%"}
```{r}
#| eval: false
#| echo: true
# YOUR TURN! Complete the below
# dplyr pipeline to group by and
# summarize the data set
library(dplyr)
data(package = "ggplot2", "diamonds")
diamonds |>
<your code here>
```
:::
::: {.column width="50%"}
```{r}
#| eval: false
#| echo: true
#| code-fold: true
#| code-summary: "See {countdown} code"
library(countdown)
countdown(
minutes = 0,
seconds = 10,
play_sound = TRUE,
color_border = "#1A4064",
color_text = "#76AADB",
font_size = "3em"
)
```
:::
:::
```{r}
#| eval: true
#| echo: false
library(countdown)
countdown(
minutes = 0,
seconds = 10,
right = "25%",
bottom = "5%",
play_sound = TRUE,
color_border = "#1A4064",
color_text = "#76AADB",
font_size = "3em"
)
```
::: footer
Quarto | Learn more about the [countdown](https://pkg.garrickadenbuie.com/countdown/) package
:::
# 2. Arrow
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
---
## Larger-than-Memory Data
<br>
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
## Arrow and Parquet `r hexes("arrow")`
- Columnar memory format for flat data
- Ultra-fast read times from Parquet files
- Easily convert data.frames and tibbles to Arrow tables in-line
- Plays well with Pandas (Python) and dplyr (R)
- Arrow will soon be able to process nested list data
::: notes
Libraries are available for C, C++, C#, Go, Java, JavaScript, Julia, MATLAB, Python, R, Ruby, and Rust
Arrow is organized for efficient analytic operations on modern hardware like CPUs and GPUs.
The Arrow memory format also supports zero-copy reads for lightning-fast data access without serialization overhead
:::
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
## How fast is fast? `r hexes("arrow")`
::: {.incremental}
- My personal laptop has 24 GB RAM
- To test Arrow's capabilities, I read a [40 GB dataset]{.fragment .highlight-teal} with over [1.1 billion rows]{.fragment .highlight-teal} and 24 columns
- The `.parquet` dataset was partitioned by Year and Month (120 files)
- Important to note that my laptop would not be able to load this object entirely into memory as a data.frame or tibble given my laptop's limited RAM
:::
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
## Benchmarking Read Times `r hexes("arrow", "ggplot2")`
::: {.codewindow .r}
benchmark_arrow_read.R
{{< include codewindow_files/benchmark_arrow_read.qmd >}}
:::
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
## Benchmarking Read Times `r hexes("arrow", "ggplot2")`
- Results show read times from a 40GB parquet, 1.1 billion row dataset (benchmarked over 1,000 iterations)
{width=55%}
::: footer
Arrow | Learn more at [https://arrow.apache.org/](https://arrow.apache.org/)
:::
## Benchmarking dplyr `r hexes("arrow", "dplyr", "ggplot2")`
::: {.codewindow .r}
benchmark_arrow_dplyr.R
{{< include codewindow_files/benchmark_arrow_dplyr.qmd >}}
:::
::: footer
Arrow | [Arrow + dplyr compatibility](https://r4ds.hadley.nz/arrow.html)
:::
## Benchmarking dplyr `r hexes("arrow", "dplyr", "ggplot2")`
- Arrow + dplyr summarized 1.1 billion rows in less than 5s (benchmarked over 10 iterations) to a 10 x 4 tibble
{width=55%}
::: footer
Arrow | [Arrow + dplyr compatibility](https://r4ds.hadley.nz/arrow.html)
:::
## Tidyverse Compatibility `r hexes("arrow", "tidyverse", "stringr")`
- Many functions from the tidyverse collections of packages have 1:1 compatibility with Arrow tables
- However, sometimes you'll encounter a breaking point
- Take this `stringr::str_replace_na()` example:
```{r}
#| echo: true
#| code-line-numbers: "1-4|3-4"
nyc_taxi |>
mutate(vendor_name = str_replace_na(vendor_name, "No vendor"))
#> Error: Expression str_replace_na(vendor_name, "No vendor")
#> not supported in Arrow
```
. . .
- This `stringr` function is not supported by Arrow
::: footer
Arrow | [Arrow + dplyr compatibility](https://r4ds.hadley.nz/arrow.html)
:::
---
---
::: r-fit-text
but wait!
:::
::: r-fit-text
problem solved
:::
---
## User Defined Functions `r hexes("arrow", "tidyverse", "dplyr", "stringr")`
- Lucky for us, Arrow allows us to create and register _**User Defined Functions ("UDFs")**_ to the Arrow engine
- Almost any function can be made compatible with Arrow by registering custom UDFs
- Let's learn how to register `str_replace_na()` with the Arrow kernel
::: footer
Arrow | Learn more about registering Arrow [User Defined Functions ("UDFs")](https://arrow.apache.org/docs/r/reference/register_scalar_function.html)
:::
## Registering UDFs `r hexes("arrow", "dplyr", "stringr", "tidyverse")`
- First, run `arrow::schema()` on your Arrow table to review the field name / data type pairs
- Since I want to mutate the `vendor_name` field, I know I'll be working with an Arrow `string()` data type
```{r}
#| echo: true
#| code-line-numbers: "1-10|3"
arrow::schema(nyc_taxi)
#> Schema
#> vendor_name: string
#> pickup_datetime: timestamp[ms]
#> dropoff_datetime: timestamp[ms]
#> passenger_count: int64
#> trip_distance: double
#> pickup_longitude: double
#> pickup_latitude: double
#> ...
```
::: footer
Arrow | Learn more about registering Arrow [User Defined Functions ("UDFs")](https://arrow.apache.org/docs/r/reference/register_scalar_function.html)
:::
## Registering UDFs `r hexes("arrow", "dplyr", "stringr", "tidyverse")`
- Next, use `register_scalar_function()`
- Name your UDF "replace_arrow_nas" and remember to set `auto_convert = TRUE`
```{r}
#| echo: true
#| code-line-numbers: "1-13|3-4|3-4,12"
arrow::register_scalar_function(
name = "replace_arrow_nas",
# Note: The first argument must always be context
function(context, x, replacement) {
stringr::str_replace_na(x, replacement)
},
in_type = schema(
x = string(),
replacement = string()
),
out_type = string(),
auto_convert = TRUE
)
```
::: notes
Unless you're developing a package, `auto_convert` should be set to `TRUE`
:::
::: footer
Arrow | Learn more about registering Arrow [User Defined Functions ("UDFs")](https://arrow.apache.org/docs/r/reference/register_scalar_function.html)
:::
## Registering UDFs `r hexes("arrow", "dplyr", "stringr", "tidyverse")`
- Try your new registered function
```{r}
#| echo: true
nyc_taxi |>
mutate(vendor_name = replace_arrow_nas(vendor_name, "No vendor")) |>
distinct(vendor_name) |>
arrange(vendor_name) |>
collect()
#> # A tibble: 3 × 1
#> vendor_name
#> <chr>
#> 1 CMT
#> 2 No vendor
#> 3 VTS
```
::: footer
Arrow | Learn more about registering Arrow [User Defined Functions ("UDFs")](https://arrow.apache.org/docs/r/reference/register_scalar_function.html)
:::
---
::: footer
Eevee can't believe it... "Wooooooooooow, Javi!"
:::
## What's next for Arrow? `r hexes("arrow")`
**_ADBC: Arrow Database Connectivity_**
- Competitor to JDBC & ODBC allowing applications to code to this API standard but fetching results in an Arrow format
::: footer
ADBC (Arrow Database Connectivity) | Learn more at [https://arrow.apache.org/docs/format/ADBC.html](https://arrow.apache.org/docs/format/ADBC.html)
:::
## What's next for Arrow? `r hexes("arrow")`
**_ADBC: Arrow Database Connectivity_**
{width=50%}
::: footer
ADBC (Arrow Database Connectivity) | Learn more at [https://arrow.apache.org/docs/format/ADBC.html](https://arrow.apache.org/docs/format/ADBC.html)
:::
## What's next for Arrow? `r hexes("arrow")`
**_Arrow Flight SQL_**
- A protocol for interacting with SQL databases using the Arrow in-memory format and the [Flight RPC](https://arrow.apache.org/docs/format/Flight.html) framework
- Its natural mode is to stream sequences of Arrow "record batches” to reduce or remove the serialization cost associated with data transport
- The design goal for Flight is to create a new protocol for data services that uses the Arrow columnar format as both the over-the-wire data representation as well as the public API presented to developers
::: footer
Arrow Flight SQL | Learn more at [https://arrow.apache.org/docs/format/FlightSql.html](https://arrow.apache.org/docs/format/FlightSql.html)
:::
# 3. DuckDB + duckplyr
::: footer
duckplyr by DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
---
## DuckDB `r hexes_duckdb("duckdb")`
- [DuckDB Labs](https://duckdblabs.com/) created an in-line database management system, like a SQLite database engine, but optimized for distributed compute and optimized for larger-than-memory analysis
- The `duckdb` package for Python and offers a state-of-the-art optimizer that pushes down filters and projections directly into Arrow scans
- As a result, only relevant columns and partitions will be read thus significantly accelerates query execution
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## DuckDB setup `r hexes_duckdb("duckdb")`
<br><br>
::: columns
::: {.column width="50%"}
```{r}
#| eval: false
#| echo: true
# R Install
install.packages("duckdb")
install.packages("arrow")
# For later use by Python and R:
arrow::copy_files(
from = "s3://ursa-labs-taxi-data",
to = here::here("nyc-taxi")
)
```
:::
::: {.column width="50%"}
```{python}
#| eval: false
#| echo: true
# Python Install
pip install duckdb
pip install pyarrow
```
:::
:::
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## DuckDB Basics `r hexes_duckdb("duckdb")`
::: {.panel-tabset}
## R
```{r}
#| eval: false
#| echo: true
library(duckdb)
library(arrow)
library(dplyr)
ds <- arrow::open_dataset("nyc-taxi", partitioning = c("year", "month"))
ds |>
filter(year > 2014 & passenger_count > 0 &
trip_distance > 0.25 & fare_amount > 0) |>
# Pass off to DuckDB
to_duckdb() |>
group_by(passenger_count) |>
mutate(tip_pct = tip_amount / fare_amount) |>
summarise(fare_amount = mean(fare_amount, na.rm = TRUE),
tip_amount = mean(tip_amount, na.rm = TRUE),
tip_pct = mean(tip_pct, na.rm = TRUE)) |>
arrange(passenger_count) |>
collect()
```
## Python
```{python}
#| eval: false
#| echo: true
import duckdb
import pyarrow as pa
import pyarrow.dataset as ds
# Open dataset using year,month folder partition
nyc = ds.dataset('nyc-taxi/', partitioning=["year", "month"])
# We transform the nyc dataset into a DuckDB relation
nyc = duckdb.arrow(nyc)
# Run same query again
nyc.filter("year > 2014 & passenger_count > 0 & trip_distance > 0.25 & fare_amount > 0")
.aggregate("SELECT AVG(fare_amount), AVG(tip_amount), AVG(tip_amount / fare_amount) as tip_pct","passenger_count").arrow()
```
:::
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## DuckDB Streaming `r hexes_duckdb("duckdb")`
::: {.panel-tabset}
## R
```{r}
#| eval: false
#| echo: true
# Reads dataset partitioning it in year/month folder
nyc_dataset = open_dataset("nyc-taxi/", partitioning = c("year", "month"))
# Gets Database Connection
con <- dbConnect(duckdb::duckdb())
# We can use the same function as before to register our arrow dataset
duckdb::duckdb_register_arrow(con, "nyc", nyc_dataset)
res <- dbSendQuery(con, "SELECT * FROM nyc", arrow = TRUE)
# DuckDB's queries can now produce a Record Batch Reader
record_batch_reader <- duckdb::duckdb_fetch_record_batch(res)
# Which means we can stream the whole query per batch.
# This retrieves the first batch
cur_batch <- record_batch_reader$read_next_batch()
```
## Python
```{python}
#| eval: false
#| echo: true
# Reads dataset partitioning it in year/month folder
nyc_dataset = ds.dataset('nyc-taxi/', partitioning=["year", "month"])
# Gets Database Connection
con = duckdb.connect()
query = con.execute("SELECT * FROM nyc_dataset")
# DuckDB's queries can now produce a Record Batch Reader
record_batch_reader = query.fetch_record_batch()
# Which means we can stream the whole query per batch.
# This retrieves the first batch
chunk = record_batch_reader.read_next_batch()
```
:::
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## DuckDB Streaming Speed `r hexes_duckdb("duckdb")`
::: {.panel-tabset}
## DuckDB
```{r}
#| eval: false
#| echo: true
# DuckDB via Python
# Open dataset using year,month folder partition
nyc = ds.dataset('nyc-taxi/', partitioning=["year", "month"])
# Get database connection
con = duckdb.connect()
# Run query that selects part of the data
query = con.execute("SELECT total_amount, passenger_count,year FROM nyc where total_amount > 100 and year > 2014")
# Create Record Batch Reader from Query Result.
# "fetch_record_batch()" also accepts an extra parameter related to the desired produced chunk size.
record_batch_reader = query.fetch_record_batch()
# Retrieve all batch chunks
chunk = record_batch_reader.read_next_batch()
while len(chunk) > 0:
chunk = record_batch_reader.read_next_batch()
```
## Pandas
```{python}
#| eval: false
#| echo: true
# We must exclude one of the columns of the NYC dataset due to an unimplemented cast in Arrow
working_columns = ["vendor_id","pickup_at","dropoff_at","passenger_count","trip_distance","pickup_longitude",
"pickup_latitude","store_and_fwd_flag","dropoff_longitude","dropoff_latitude","payment_type",
"fare_amount","extra","mta_tax","tip_amount","tolls_amount","total_amount","year", "month"]
# Open dataset using year,month folder partition
nyc_dataset = ds.dataset(dir, partitioning=["year", "month"])
# Generate a scanner to skip problematic column
dataset_scanner = nyc_dataset.scanner(columns=working_columns)
# Materialize dataset to an Arrow Table
nyc_table = dataset_scanner.to_table()
# Generate Dataframe from Arow Table
nyc_df = nyc_table.to_pandas()
# Apply Filter
filtered_df = nyc_df[
(nyc_df.total_amount > 100) &
(nyc_df.year >2014)]
# Apply Projection
res = filtered_df[["total_amount", "passenger_count","year"]]
# Transform Result back to an Arrow Table
new_table = pa.Table.from_pandas(res)
```
:::
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## DuckDB Streaming Speed `r hexes_duckdb("duckdb")`
- Pandas runtime was 146.91 seconds
. . .
- DuckDB runtime was [0.05 seconds]{.fragment .highlight-teal}
. . .
{width=60%}
::: footer
DuckDB Labs | Learn how [DuckDB quacks Arrow](https://duckdb.org/2021/12/03/duck-arrow.html)
:::
## duckplyr `r hexes_duckdb("duckdb")`
- `duckplyr`, from DuckDB Labs, offers 1:1 compatibility with `dplyr` functions but there are some caveats:
- factor columns, nested lists, and nested tibbles are not _yet_ supported
- you have to use `.by` in `dplyr::summarize()` as `dplyr::group_by()` will not be supported by the developers
::: footer
duckplyr | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
## Benchmarking Analysis `r hexes_duckdb("duckdb")`
::: columns
::: {.column width="43%"}
- Arrow and DuckDB really stood out for fast manipulation of data using `dplyr` syntax
:::
::: {.column width="57%"}
- The code below shows the basic transformation done to the NYC Taxi dataset via `dplyr`, `arrow`, `duckdb`, and `duckplyr`
:::
:::
```{r}
#| eval: false
#| echo: true
#| code-line-numbers: "1-7|7"
nyc_taxi |>
dplyr::filter(passenger_count > 1) |>
dplyr::group_by(year) |>
dplyr::summarise(all_trips = n(),
shared_trips = sum(passenger_count, na.rm = T)) |>
dplyr::mutate(pct_shared = shared_trips / all_trips * 100) |>
dplyr::collect()
```
::: footer
DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
## Benchmark: 1 million rows `r hexes_duckdb("duckdb")`
::: footer
DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
## Benchmark: 10 million rows `r hexes_duckdb("duckdb")`
::: footer
DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
## Benchmark: 100 million rows `r hexes_duckdb("duckdb")`
::: footer
DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
## Benchmark: 500 million rows `r hexes_duckdb("duckdb")`
::: footer
DuckDB Labs | Learn more at [https://duckdblabs.com/](https://duckdblabs.com/)
:::
# 4. tidymodels
::: footer
tidymodels | Learn more at [https://www.tidymodels.org/](https://www.tidymodels.org/)
:::
## {background-image="images/tm-org.png" background-size="contain"}
::: footer
tidymodels | Learn more at [https://www.tidymodels.org/](https://www.tidymodels.org/)
:::
## tidymodels packages `r hexes("tidymodels", "rsample", "parsnip")`
{width=75%}
::: footer
tidymodels | Learn more at [https://www.tidymodels.org/](https://www.tidymodels.org/)
:::
## tidymodels packages `r hexes("recipes", "workflows", "tune")`
{width=75%}
::: footer
tidymodels | Learn more at [https://www.tidymodels.org/](https://www.tidymodels.org/)
:::
## tidymodels packages `r hexes("yardstick", "broom", "dials")`
{width=75%}
::: footer
tidymodels | Learn more at [https://www.tidymodels.org/](https://www.tidymodels.org/)
:::
## usemodels `r hexes("tidymodels")`
- The `usemodels` package generates easy-to-use boilerplate code for modeling
- `usemodels::use_glmnet()` generates code that eases the setup of using the [glmnet](https://glmnet.stanford.edu/index.html) algorithm, _and others_, within the `tidymodels` framework
### usemodels input example
```{r}
#| eval: false
#| echo: true
#| code-line-numbers: "1-4|4"
library(usemodels)
library(palmerpenguins)
data(penguins)
use_glmnet(body_mass_g ~ ., data = penguins)
```
::: footer
usemodels | Learn more at [https://usemodels.tidymodels.org/](https://usemodels.tidymodels.org/)
:::
## usemodels output `r hexes("tidymodels")`
```{r}
#| eval: false
#| echo: true
glmnet_recipe <- recipe(formula = body_mass_g ~ ., data = penguins) |>
step_zv(all_predictors()) |>
step_normalize(all_numeric_predictors())
glmnet_spec <- multinom_reg(penalty = tune(), mixture = tune()) |>
set_mode("classification") |>
set_engine("glmnet")
glmnet_workflow <- workflow() |>
add_recipe(glmnet_recipe) |>
add_model(glmnet_spec)
glmnet_grid <- tidyr::crossing(penalty = 10^seq(-6, -1, length.out = 20),
mixture = c(0.05, 0.2, 0.4, 0.6, 0.8, 1))
glmnet_tune <-
tune_grid(glmnet_workflow,
resamples = stop("add your rsample object"),
grid = glmnet_grid)
```
::: footer
usemodels | Learn more at [https://usemodels.tidymodels.org/](https://usemodels.tidymodels.org/)
:::
## usemodels templates `r hexes("tidymodels")`
- `usemodels` isn't new but still in early development
- the following model templates can be called using one of the `usemodels::use_*()` variants below:
```{r}
#| eval: false
#| echo: true
> ls("package:usemodels", pattern = "use_")
[1] "use_C5.0" "use_cubist" "use_earth"
[4] "use_glmnet" "use_kernlab_svm_poly" "use_kernlab_svm_rbf"
[7] "use_kknn" "use_ranger" "use_xgboost"
```
::: footer
usemodels | Learn more at [https://usemodels.tidymodels.org/](https://usemodels.tidymodels.org/)
:::
## Hyperparameter Tuning `r hexes("dials", "tune")`
The main two strategies for optimization are:
. . .
- **Grid search** 💠 which tests a pre-defined set of candidate values
- **Iterative search** 🌀 which suggests/estimates new values of candidate parameters to evaluate
::: footer
Slide credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Grid Search `r hexes("tune")`
- In this example, we see a large grid showing all points with our goal being to minimize MAE via learning rate:
{width=55%}
::: footer
Slide & illustration credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Grid Search `r hexes("tune")`
- In reality we would probably sample the space more densely with, for example, a three point grid:
{width=55%}
::: footer
Slide & illustration credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Iterative Search `r hexes("tune")`
- We could start with a few points and search the space using an iterative approach via the `tune` package:
::: footer
Slide & illustration credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Gaussian Process & Optimization
We can make a “meta-model” with a small set of historical performance results
_Gaussian Processes (GP)_ models are a good choice to model performance
- It is a Bayesian model so we are using _Bayesian Optimization (BO)_
- The GP model can take candidate tuning parameter combinations as inputs and make predictions for performance (e.g. MAE)
::: footer
Slide credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Acquisition Functions
We’ll use an acquisition function to select a new candidate
The most popular method appears to be _Expected Improvement (EI)_ above the current best results
- Zero at existing data points.
- The expected improvement is integrated over all possible improvement (“expected” in the probability sense).
We would probably pick the point with the largest EI as the next point to iterate
::: footer
Slide credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Iteration Approach
Once we pick the candidate point, we measure performance for it (e.g. resampling)
Another GP is fit, EI is recomputed, and so on
We stop when we have completed the allowed number of iterations or if we don’t see any improvement after a pre-set number of attempts
::: footer
Slide credit | [Posit, Advanced Tidymodels Workshop](https://workshops.tidymodels.org/#advanced-tidymodels)
:::
## Iteration & EI Visualized
