ianmcook commented on a change in pull request #158:
URL: https://github.com/apache/arrow-site/pull/158#discussion_r744832926
##########
File path: _posts/2021-11-05-r-6.0.0.md
##########
@@ -0,0 +1,235 @@
+---
+layout: post
+title: Apache Arrow R 6.0.0 Release
+date: "2021-11-08"
+author: Nic Crane, Jonathan Keane, Neal Richardson
+categories: [release]
+---
+<!--
+{% comment %}
+Licensed to the Apache Software Foundation (ASF) under one or more
+contributor license agreements. See the NOTICE file distributed with
+this work for additional information regarding copyright ownership.
+The ASF licenses this file to you under the Apache License, Version 2.0
+(the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+{% endcomment %}
+-->
+
+We are excited to announce the recent release of version 6.0.0 of the Arrow R
package on [CRAN](https://cran.r-project.org/package=arrow). While we usually
don't write a dedicated release blog post for the R package, this one is
special. There are a number of major new features in this version, some of
which we've been building up to for several years.
+
+# More dplyr support
+
+In version 0.16.0 (February 2020), we released the first version of the
Dataset feature, which allowed you to query multi-file datasets using
`dplyr::select()` and `filter()`. These tools allowed you to find a slice of
data in a large dataset that may not fit into memory and pull it into R for
further analysis. In version 4.0.0 earlier this year, we added support for
`mutate()` and a number of other dplyr verbs, and all year we've been adding
hundreds of functions you can use to transform and filter data in Datasets.
However, to aggregate, you'd still need to pull the data into R.
+
+## Grouped aggregation
+
+With `arrow` 6.0.0, you can now `summarise()` on Arrow data, both with or
without `group_by()`. These are supported both with in-memory Arrow tables as
well as across partitioned datasets. Most common aggregation functions are
supported: `n()`, `n_distinct()`, `min(),` `max()`, `sum()`, `mean()`, `var()`,
`sd()`, `any()`, and `all()`. `median()` and `quantile()` with one probability
are also supported and currently return approximate results using the t-digest
algorithm.
+
+As usual, Arrow will read and process data in chunks and in parallel when
possible to produce results much faster than one could by loading it all into
memory then processing. This allows for operations that wouldn't fit into
memory on a single machine. For example, using the 1.5-billion row NYC Taxi
dataset we use for examples in the [package
vignette](https://arrow.apache.org/docs/r/articles/dataset.html), we can
aggregate over the whole dataset even on a laptop:
+
+```r
+ds <- open_dataset("nyc-taxi", partitioning = c("year", "month"))
+ds %>%
+ filter(
+ passenger_count > 0,
+ passenger_count < 6,
+ grepl("csh", payment_type, ignore.case = TRUE)
+ ) %>%
+ group_by(passenger_count) %>%
+ summarize(
+ avg = mean(total_amount, na.rm = TRUE),
+ count = n()
+ ) %>%
+ arrange(desc(count)) %>%
+ collect()
+
+#> # A tibble: 5 × 3
+#> passenger_count avg count
+#> <int> <dbl> <int>
+#> 1 1 11.1 257738064
+#> 2 2 12.1 58824482
+#> 3 5 11.4 26056438
+#> 4 3 12.0 18852606
+#> 5 4 12.3 10081632
+```
+
+## Joins
+
+In addition to aggregation, Arrow also supports all of dplyr's mutating joins
(inner, left, right, and full) and filtering joins (semi and anti).
+
+Suppose I want to get a table of all the flights from JFK to Las Vegas Airport
on
+9th October 2013, with the full name of the airline included.
+
+```r
+arrow_table(nycflights13::flights) %>%
+ filter(
+ year == 2013,
+ month == 10,
+ day == 9,
+ origin == "JFK",
+ dest == "LAS"
+ ) %>%
+ select(dep_time, arr_time, carrier) %>%
+ left_join(
+ arrow_table(nycflights13::airlines)
+ ) %>%
+ collect()
+
+#> # A tibble: 12 × 4
+#> dep_time arr_time carrier name
+#> <int> <int> <chr> <chr>
+#> 1 637 853 B6 JetBlue Airways
+#> 2 648 912 AA American Airlines Inc.
+#> 3 812 1029 DL Delta Air Lines Inc.
+#> 4 945 1206 VX Virgin America
+#> 5 955 1219 B6 JetBlue Airways
+#> 6 1018 1231 DL Delta Air Lines Inc.
+#> 7 1120 1338 B6 JetBlue Airways
+#> 8 1451 1705 DL Delta Air Lines Inc.
+#> 9 1656 1915 AA American Airlines Inc.
+#> 10 1755 2001 DL Delta Air Lines Inc.
+#> 11 1827 2049 B6 JetBlue Airways
+#> 12 1917 2126 DL Delta Air Lines Inc.
+```
+
+In this example, we're working on an in-memory table, so you wouldn't need
`arrow` to do this--but the same code would work on a larger-than-memory
dataset backed by thousands of Parquet files.
+
+## Under the hood
+
+To support these features, we've made some internal changes to how queries are
built up and--importantly--when they are evaluated. As a result, there are some
changes in behavior compared to past versions of `arrow`.
+
+First, calls to `summarise()`, `head()`, and `tail()` no longer eagerly
evaluate: this means you need to call either `compute()` (to evaluate it and
produce an Arrow Table) or `collect()` (to evaluate and pull the Table into an
R `data.frame`) to see the results.
+
+Second, the order of rows in a dataset query is no longer determinisitic due
to the way the parallelization of work happens in the C++ library. This means
that you can't assume that the results of a query will be in the same order as
the rows of data in the files on disk. If you do need a stable sort order, call
`arrange()` to specify ordering.
+
+While these changes are a break from past `arrow` behavior, they are
consistent with many `dbplyr` backends and are needed to allow queries to scale
beyond data-frame workflows that can fit into memory.
+
+# Integration with DuckDB
+
+The Arrow engine is not the only new way to query Arrow Datasets in this
release. If you have the [duckdb](https://duckdb.org/) package installed, you
can hand off an Arrow Dataset or query object to duckdb for further querying
using the `to_duckdb()` function. This allows you to use duckdb's `dbplyr`
methods, as well as its SQL interface, to aggregate data. DuckDB supports
filter pushdown, so you can take advantage of Arrow Datasets and Arrow-based
optimizations even within a DuckDB SQL query with a `where` clause. Filtering
and column projection specified before the `to_duckdb()` call in a pipeline is
evaluated in Arrow; this can be helpful in some circumstances like complicated
dbplyr pipelines. You can also hand off DuckDB data (or the result of a query)
to arrow with the `to_arrow()` call.
Review comment:
```suggestion
The Arrow engine is not the only new way to query Arrow Datasets in this
release. If you have the [duckdb](https://cran.r-project.org/package=duckdb)
package installed, you can hand off an Arrow Dataset or query object to
[DuckDB](https://duckdb.org/) for further querying using the `to_duckdb()`
function. This allows you to use duckdb's `dbplyr` methods, as well as its SQL
interface, to aggregate data. DuckDB supports filter pushdown, so you can take
advantage of Arrow Datasets and Arrow-based optimizations even within a DuckDB
SQL query with a `where` clause. Filtering and column projection specified
before the `to_duckdb()` call in a pipeline is evaluated in Arrow; this can be
helpful in some circumstances like complicated dbplyr pipelines. You can also
hand off DuckDB data (or the result of a query) to arrow with the `to_arrow()`
call.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
