We have been writing a lot on higher-order data transforms lately:
- Coordinatized Data: A Fluid Data Specification
- Data Wrangling at Scale
- Fluid Data
- Big Data Transforms.
What I want to do now is "write a bit more, so I finally feel I have been concise."
R package supplies general data transform operators.
- The whole system is based on two primitives or operators
- These operators have pivot, un-pivot, one-hot encode, transpose, moving multiple rows and columns, and many other transforms as simple special cases.
- It is easy to write many different operations in terms of the
- These operators can work-in memory or at big data scale (with databases and Apache Spark; for big data we use the
- The transforms are controlled by a control table that itself is a diagram of or picture of the transform.
We will end with a quick example, centered on pivoting/un-pivoting values to/from more than one column at the same time.
Suppose we had some sales data supplied as the following table:
Suppose we are interested in adding a derived column: which region the salesperson made most of their bookings in.
## Loading required package: wrapr
d <- d %.>% dplyr::mutate(., BestRegion = ifelse(BookingsWest > BookingsEast, "West", ifelse(BookingsEast > BookingsWest, "East", "Both")))
Our notional goal is (as part of a larger data processing plan) to reformat the data a thin/tall table or a RDF-triple like form. Further suppose we wanted to copy the derived column into every row of the transformed table (perhaps to make some other step involving this value easy).
We can use
cdata::moveValuesToRowsD() to do this quickly and easily.
First we design what is called a transform control table.
cT1 <- data.frame(Region = c("West", "East"), Bookings = c("BookingsWest", "BookingsEast"), BestRegion = c("BestRegion", "BestRegion"), stringsAsFactors = FALSE) print(cT1)
## Region Bookings BestRegion ## 1 West BookingsWest BestRegion ## 2 East BookingsEast BestRegion
In a control table:
- The column names specify new columns that will be formed by
- The values specify where to take values from.
This control table is called "non trivial" as it does not correspond to a simple pivot/un-pivot (those tables all have two columns). The control table is a picture of of the mapping we want to perform.
An interesting fact is
cdata::moveValuesToColumnsD(cT1, cT1, keyColumns = NULL) is a picture of the control table as a one-row table (and this one row table can be mapped back to the original control table by
cdata::moveValuesToRowsD(), these two operators work roughly as inverses of each other; though
cdata::moveValuesToRowsD() operates on rows and
cdata::moveValuesToColumnsD() operates on groups of rows specified by the keying columns).
The mnemonic is:
cdata::moveValuesToColumnsD()converts arbitrary grouped blocks of rows that look like the control table into many columns.
cdata::moveValuesToRowsD()converts each row into row blocks that have the same shape as the control table.
Because pivot and un-pivot are fairly common needs
cdata also supplies functions that pre-populate the controls tables for these operations (
To design any transform you draw out the control table and then apply one of these operators (you can pretty much move from any block structure to any block structure by chaining two or more of these steps).
We can now use the control table to supply the same transform for each row.
d %.>% dplyr::mutate(., Quarter = substr(Period,5,6), Year = as.numeric(substr(Period,1,4))) %.>% dplyr::select(., -Period) %.>% moveValuesToRowsD(., controlTable = cT1, columnsToCopy = c('SalesPerson', 'Year', 'Quarter')) %.>% arrange_se(., c('SalesPerson', 'Year', 'Quarter', 'Region')) %.>% knitr::kable(.)
Notice we were able to easily copy the extra
BestRegion values into all the correct rows.
It can be hard to figure out how to specify such a transformation in terms of pivots and un-pivots. However, as we have said: by drawing control tables one can easily design and manage fairly arbitrary data transform sequences (often stepping through either a denormalized intermediate where all values per-instance are in a single row, or a thin intermediate like the triple-like structure we just moved into).
Categories: Pragmatic Data Science Tutorials
Data Scientist and trainer at Win Vector LLC. One of the authors of Practical Data Science with R.
The need to move multiple columns in a pivot or un-pivot shows up a lot in practice, it just tends to be under-served by packages. That is why I thought it was a good example to show some of the ways cdata stands out.
Could you explain when this approach might be preferred over tidyr’s spread and gather functions? Is the main idea to cover big-data table backends, which spread and gather don’t (yet) cover?
Good question, and thanks!
In my opinion
cdata‘s primary advantages are:
cdataprovides the ability to carry multiple columns in either direction, meaning it has easy access to many more transforms that just
cdataworks on big-data back ends (in particular
Apache Spark, in addition to in-memory data. I don’t know if
tidyrhas big data back ends on its schedule, so “yet” may or may not be the right way to view
unpivotValuesToRows(currently only available for in-memory processing) check and maintain more invariants than
tidyr::spread(). This allows one to ensure and document a sequence of these operations is safely composable and reversible.
(Controversial one) I think the standard evaluation interfaces are better (easier to program over) than the non-standard column naming schemes used in
tidyr. Also given my experience in moving from
dplyr 0.7.*I expect a lot of breakage in
tidyrusing code when (as I assume it will)
tidyrswitches from its own interface implementation to
(Controversial one) I have found
unpivotValuesToRowseasier to teach.
cdatais newer code with
much less experience in production than
is supplied by
RSQLite, so any problems arising from this bridge (dates, list columns, S4 class columns) are going to be present in
So to as to when
cdatais better: I think quite often. Our group is not developing any new code on top of
tidyrunless clients specifically request it.
tidyrauthors will have a “similar opinion with some signs reversed” (i.e., prefer their own work).