Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Vignetes: Improve linking, indentation and respect 120 character limit #155

Merged
merged 5 commits into from
Oct 4, 2024
Merged

Vignetes: Improve linking, indentation and respect 120 character limit #155

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
9a326d0
docs(basic-principles): Improve linking and indentation
RasmusSkytte Mar 4, 2024
19fdb52
docs(basic-principles): Remove early Note
RasmusSkytte Mar 4, 2024
7de23de
docs(slowly-changing-dimensions): Respect 120 character limit
RasmusSkytte Oct 4, 2024
9b76199
docs(benchmarks): Improve linking
RasmusSkytte Oct 4, 2024
a9293c3
fix(basic-principles): run `get_connection()` only once
RasmusSkytte Oct 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
65 changes: 39 additions & 26 deletions vignettes/SCDB.Rmd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,8 @@ example_data <-
name = "example_data",
overwrite = TRUE)
```
The basic principle of the SCDB package is to enable the user to easily implement and maintain a database of time-versioned data.
The basic principle of the SCDB package is to enable the user to easily implement and maintain a database of
time-versioned data.

In practice, this is done by labeling each record in the data with three additional fields:

Expand All @@ -47,59 +48,69 @@ In practice, this is done by labeling each record in the data with three additio

This strategy of time versioning is often called "type 2" history [@Kimball2013].

Note that identical records may be removed and introduced more than once; for example, in a table of names and addresses, a person may change their address (or name) back to a previous value.

The SCDB package provides the function `update_snapshot` to handle the insertion and deactivation of records using this strategy.
The SCDB package provides the function `update_snapshot` to handle the insertion and deactivation of records using
this strategy.
It further includes several functions to improve the Quality of life for working with database data.

A simple example of usage is shown below.<br/>
For this example, we use a temporary, on-disk SQLite database.
Note that `get_connection` tries to establish connection using `DBI::dbConnect` with as few additional arguments as possible.
Note that `get_connection()` tries to establish connection using `DBI::dbConnect()` with as few additional arguments
as possible.
Different drivers may require authentication which can be read from a configuration file.^[
In the context of the SCDB package, this is most notably `RPostgres::Postgres()`, which may read from a `.pgpass` file.
See also the [PostgreSQL documentation](https://www.postgresql.org/docs/current/libpq-pgpass.html).
]

Our example data is `datasets::mtcars` reduced to only two columns: row names converted to a column `car`, and `hp`
```{r example_data, eval = requireNamespace("RSQLite", quietly = TRUE)}
if (!exists("conn")) conn <- get_connection()

```{r example_data, eval = FALSE}
conn <- get_connection()
```
```{r example_data_hidden, eval = requireNamespace("RSQLite", quietly = TRUE)}
example_data <- dplyr::tbl(conn, DBI::Id(table = "example_data"))
example_data
```

Imagine on Day 1, in this case January 1st, 2020, our currently available data is the first three records of the example_data.
Imagine on Day 1, in this case January 1st, 2020, our currently available data is the first three records of
the `example_data`.
We then store this data in a table `mtcars`:
```{r example_1, eval = requireNamespace("RSQLite", quietly = TRUE)}
data <- head(example_data, 3)

update_snapshot(.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-01 11:00:00"))
update_snapshot(
.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-01 11:00:00")
)
```

We can then access out data using the `get_table` function, and include information on data validity period using `include_slice_info = TRUE`:
We can then access out data using the `get_table()` function, and include information on data validity period using
`include_slice_info = TRUE`:
```{r example_1_results, eval = requireNamespace("RSQLite", quietly = TRUE)}
get_table(conn, "mtcars")

get_table(conn, "mtcars", include_slice_info = TRUE)
```
Note that where e.g. `dplyr::tbl` requires a more exact specification of the table identity (`tbl(conn, DBI::Id(table = "mtcars"))`), `get_table` will parse any character to a `DBI::Id` object input using `SCDB::id`.
Note that where e.g. `dplyr::tbl()` requires a more exact specification of the table identity
(`tbl(conn, DBI::Id(table = "mtcars"))`), `get_table()` will parse any character to a `DBI::Id()` object input
using `id()`.

The following day, the current data is now the first five rows of our example data.
We then store this data in the database using update_snapshot:
We then store this data in the database using `update_snapshot()`:

```{r example_2, eval = requireNamespace("RSQLite", quietly = TRUE)}
# Let's say that the next day, our data set is now the first 5 of our example data
data <- head(example_data, 5)

update_snapshot(.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-02 12:00:00"))
update_snapshot(
.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-02 12:00:00")
)
```
We can again use the `get_table` function to see the latest available data, including time-keeping with `include_slice_info = TRUE`:
We can again use the `get_table()` function to see the latest available data, including time-keeping with
`include_slice_info = TRUE`:
```{r example_2_results_a, eval = requireNamespace("RSQLite", quietly = TRUE)}
get_table(conn, "mtcars")

Expand All @@ -117,13 +128,15 @@ On day 3, we imagine that we have the same 5 records, but one of them is altered
data <- head(example_data, 5) |>
dplyr::mutate(hp = ifelse(car == "Mazda RX4", hp / 2, hp))

update_snapshot(.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-03 10:00:00"))
update_snapshot(
.data = data,
conn = conn,
db_table = "mtcars", # the name of the DB table to store the data in
timestamp = as.POSIXct("2020-01-03 10:00:00")
)
```

We can again access our data using the `get_table` function and see that the currently
We can again access our data using the `get_table()` function and see that the currently
available data (with the changed hp value for Mazda RX4)
```{r example_3_results_a, eval = requireNamespace("RSQLite", quietly = TRUE)}
get_table(conn, "mtcars")
Expand Down
4 changes: 2 additions & 2 deletions vignettes/benchmarks.Rmd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ This data forms the basis for three "snapshots" used in the benchmarks:
The benchmark function uses three consecutive calls to `update_snapshot()` to create the table with first snapshot and
then update it to the second and third snapshot. Finally, the table is deleted.

The performance of this benchmark function is timed with the `microbenchmark` package using 10 replicates.
The performance of this benchmark function is timed with the `{{microbenchmark}}` package using 10 replicates.
All benchmarks are run on the same machine.

The results of the benchmark are shown graphically below (mean and standard deviation), where we compare the current
Expand Down Expand Up @@ -126,7 +126,7 @@ data. The datasets are similar to the first benchmark, but the number of repeats
increasing data size. The benchmarks are run from a "clean" state, where the target_table does not exists. The benchmark
measures both the time to create the table and to remove it again afterwards (to restore the clean state).

The performance of this benchmark function is timed with the `microbenchmark` package using 5 replicates.
The performance of this benchmark function is timed with the `{{microbenchmark}}` package using 5 replicates.
All benchmarks are run on the same machine.

The results of the benchmark are shown graphically below (mean and standard deviation) and with linear scaling (dotted
Expand Down
47 changes: 32 additions & 15 deletions vignettes/slowly-changing-dimension.Rmd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,8 @@ forecasts2 <- filter(forecasts_full, ForecastDate == "2023-09-29") %>%
forecasts2a <- filter(forecasts_full, ForecastDate == "2023-09-29")
```

A slowly changing dimension is a concept in data warehousing which refers to data which may change over time, but at an irregular schedule.
A slowly changing dimension is a concept in data warehousing which refers to data which may change over time, but at an
irregular schedule.

## Type 1 and Type 2 history

Expand All @@ -64,10 +65,12 @@ For example, consider the following table of forecasts for a number of cities:
forecasts
```

The following day, the forecasts will have changed, and — barring the occasional data hoarder — the existing data is no longer relevant.
The following day, the forecasts will have changed, and — barring the occasional data hoarder — the existing data is no
longer relevant.

In this example, most (if not all) of the values of the `Forecast` column will change with each regular update.
Putting it into other words, the table is a *snapshot*^[A *snapshot* is a static view of (part of) a database at a specific point in time] of forecasts at the last time of update.
Putting it into other words, the table is a *snapshot*^[A *snapshot* is a static view of (part of) a database at a
specific point in time] of forecasts at the last time of update.

The following day, the forecasts naturally change:
```{r forecasts2a, eval = requireNamespace("tidyverse")}
Expand All @@ -77,14 +80,17 @@ forecasts2
We could choose to update the forecasts table so that it would always contain the current data.
This is what is referred to as Type 1 methodology [@Kimball2013].

Databases are thankfully a rather efficient way of storing and accessing data, so instead of discarding the values from the previous day, we append the new data to those of the previous day.
Databases are thankfully a rather efficient way of storing and accessing data, so instead of discarding the values from
the previous day, we append the new data to those of the previous day.
Also, in order to keep our data organized, we add a column with the date of the forecast, aptly named `ForecastDate`.

The full table of forecasts for the two days now looks like below, and we are slowly building a full history of forecasts:
The full table of forecasts for the two days now looks like below, and we are slowly building a full history of
forecasts:
```{r forecasts_full, eval = requireNamespace("tidyverse")}
forecasts_full
```
Managing historical data by inserting new data in this manner is often referred to as Type 2 methodology or Type 2 history.
Managing historical data by inserting new data in this manner is often referred to as Type 2 methodology or Type 2
history.

Our table now provides much more information for the user through filtering:
```{r forecasts_full_examples, eval = requireNamespace("tidyverse")}
Expand All @@ -105,7 +111,8 @@ forecasts_full %>%

Now, we note that the forecast for Houston has not changed between the two days.

In order to keep our data as minimized as possible, we modify the table again, now expanding `ForecastDate` into `ForecastFrom` and `ForecastUntil`.
In order to keep our data as minimized as possible, we modify the table again, now expanding `ForecastDate` into
`ForecastFrom` and `ForecastUntil`.

Our table of forecasts now looks like this:
```{r forecasts_scd, eval = requireNamespace("tidyverse")}
Expand All @@ -114,10 +121,12 @@ forecasts_scd
For now, the `ForecastUntil` value is set to `NA`, as it is not known when these rows will "expire" (if ever).
This also makes it easy to identify currently valid data.

Adding a new column to save a single row of data naturally seems a bit overkill, but as the number of rows in the data set increases indefinitely, this solutions scales much better.
Adding a new column to save a single row of data naturally seems a bit overkill, but as the number of rows in the data
set increases indefinitely, this solutions scales much better.

## A "timeline of timelines"
Let's now introduce additional information and see how managing slowly changing dimensions enables us to easily navigate large amounts of data over large periods of time.
Let's now introduce additional information and see how managing slowly changing dimensions enables us to easily navigate
large amounts of data over large periods of time.

```{r adresses_setup, include = FALSE, eval = requireNamespace("tidyverse")}
addresses <- tibble::tibble(
Expand Down Expand Up @@ -176,9 +185,13 @@ addresses2 <- addresses %>%
distinct()
```

Imagine a town of several thousand citizens, with a town hall maintaining a civil registry of names and addresses of every citizen, updated daily with any changes submitted by the citizens, each of whom having an individual identification number.^[If this concept seems very familiar, you may have heard of [the Danish central civil registry](https://en.wikipedia.org/wiki/Personal_identification_number_(Denmark))]
Imagine a town of several thousand citizens, with a town hall maintaining a civil registry of names and addresses of
every citizen, updated daily with any changes submitted by the citizens, each of whom having an individual
identification number.^[If this concept seems very familiar, you may have heard of
[the Danish central civil registry](https://en.wikipedia.org/wiki/Personal_identification_number_(Denmark))]

The data is largely static, as a very small fraction of citizens move on any given day, but it is of interest to keep data relatively up-to-date.
The data is largely static, as a very small fraction of citizens move on any given day, but it is of interest to keep
data relatively up-to-date.
This is where managing a slowly changing dimension becomes very powerful, compared to full incremental backups.

One day, Alice Doe meets [Robert "Bobby" Tables](https://xkcd.com/327/), and they move in together:
Expand All @@ -187,7 +200,8 @@ One day, Alice Doe meets [Robert "Bobby" Tables](https://xkcd.com/327/), and the
addresses
```

First thing to notice is that the registry is not updated in real-time, as citizens may have been late in registering a change of address.
First thing to notice is that the registry is not updated in real-time, as citizens may have been late in registering
a change of address.
This can be seen when comparing the values of `MovedIn` and `ValidFrom` for row 4.

When using Type 2 history, this feature is correctly replicated when reconstructing historical data:
Expand All @@ -200,7 +214,8 @@ addresses %>%
ValidUntil >= !!slice_timestamp | is.na(ValidUntil)) %>%
select(!c("ValidFrom", "ValidUntil"))
```
In other words, even though Alice's address was subsequently updated in the registry, we can still see that she *was* registered as living in Donut Plains at this time.
In other words, even though Alice's address was subsequently updated in the registry, we can still see that she *was*
registered as living in Donut Plains at this time.
This modeling of "timelines of timelines" is also called bitemporal modeling.

By now, things are going well between Alice and Robert; they get married, with Alice taking Robert's surname.
Expand All @@ -212,7 +227,8 @@ filter(addresses2,
select(ID, GivenName, Surname, MovedIn, MovedOut, ValidFrom, ValidUntil)
```

This is now also reflected in the data; the `MovedIn` date is persistent across the date of the name change, only the `Surname` changes:
This is now also reflected in the data; the `MovedIn` date is persistent across the date of the name change, only the
`Surname` changes:
```{r addresses3, eval = requireNamespace("tidyverse")}
slice_timestamp <- "2022-03-04"

Expand All @@ -235,7 +251,8 @@ addresses2 %>%

## Summary

By now, it is hopefully clear how managing a slowly changing dimension allows you to access data at any point in (tracked) time while potentially avoiding a lot of data redundancy.
By now, it is hopefully clear how managing a slowly changing dimension allows you to access data at any point in
(tracked) time while potentially avoiding a lot of data redundancy.

You are now ready to get started with the `SCDB` package!

Expand Down
Loading