7  Building pipelines

In previous chapters, we learned a large number of transforms for manipulating and processing geotables. In all those code examples, we used Julia’s pipe operator |> to apply the transform and send the resulting geotable to the next transform:

geotable |> transform1 |> transform2 |> ... |> viewer

In this chapter, we will learn two new powerful operators → and ⊔ provided by the framework to combine transforms into pipelines that can be optimized and reused with different geotables.

7.1 Motivation

The pipe operator |> in Julia is very convenient for sequential application of functions. Given an input x, we can type x |> f1 |> f2 to apply functions f1 and f2 in sequence, in a way that is equivalent to f2(f1(x)) or, alternatively, to the function composition (f2 ∘ f1)(x). Its syntax can drastically improve code readability when the number of functions is large. However, the operator has a major limitation in the context of geospatial data science: it evaluates all intermediate results as soon as the data is inserted in the pipe. This is known in computer science as eager evaluation.

Taking the expression above as an example, the operator will first evaluate f1(x) and store the result in a variable y. After f1 is completed, the operator evaluates f2(y) and produces the final (desired) result. If y requires a lot of computer memory as it is usually the case with large geotables, the application of the pipeline will be slow.

Another evaluation strategy, known as lazy evaluation, consists of building the entire pipeline without the data in it. The major advantage of this strategy is that it can analyze the functions, and potentially simplify the code before evaluation. For example, the pipeline cos → acos can be replaced by the much simpler pipeline identity for some values of the input x.

7.2 Operator →

In our framework, the operator → (\to) can be used in place of the pipe operator to build lazy sequential pipelines of transforms. Consider the synthetic data from previous chapters:

N = 10000
a = [2randn(N÷2) .+ 6; randn(N÷2)]
b = [3randn(N÷2); 2randn(N÷2)]
c = randn(N)
d = c .+ 0.6randn(N)

table = (; a, b, c, d)

gtb = georef(table, CartesianGrid(100, 100))
9990 rows omitted
10000×5 GeoTable over 100×100 CartesianGrid
a b c d geometry
Continuous Continuous Continuous Continuous Quadrangle
[NoUnits] [NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
5.69324 -0.318443 -0.437185 0.169292 Quadrangle((x: 0.0 m, y: 0.0 m), ..., (x: 0.0 m, y: 1.0 m))
6.59929 0.455213 1.47897 1.07508 Quadrangle((x: 1.0 m, y: 0.0 m), ..., (x: 1.0 m, y: 1.0 m))
6.69226 0.27549 0.87714 0.853007 Quadrangle((x: 2.0 m, y: 0.0 m), ..., (x: 2.0 m, y: 1.0 m))
3.67113 -0.262194 -1.08956 -1.17638 Quadrangle((x: 3.0 m, y: 0.0 m), ..., (x: 3.0 m, y: 1.0 m))
9.25026 -4.44372 -0.196566 -0.17191 Quadrangle((x: 4.0 m, y: 0.0 m), ..., (x: 4.0 m, y: 1.0 m))
9.07536 1.83108 0.0595348 -0.170243 Quadrangle((x: 5.0 m, y: 0.0 m), ..., (x: 5.0 m, y: 1.0 m))
7.53859 -1.17863 0.293477 0.41661 Quadrangle((x: 6.0 m, y: 0.0 m), ..., (x: 6.0 m, y: 1.0 m))
8.07012 5.16173 -0.953416 -0.79797 Quadrangle((x: 7.0 m, y: 0.0 m), ..., (x: 7.0 m, y: 1.0 m))
4.22035 1.72536 -0.890787 -1.74594 Quadrangle((x: 8.0 m, y: 0.0 m), ..., (x: 8.0 m, y: 1.0 m))
5.85789 -4.03981 0.880018 1.73982 Quadrangle((x: 9.0 m, y: 0.0 m), ..., (x: 9.0 m, y: 1.0 m))
⋮ ⋮ ⋮ ⋮ ⋮

And suppose that we are interested in converting the columns “a”, “b” and “c” of the geotable with the Quantile transform. Instead of creating the intermediate geotable with the Select transform, and then sending the result to the Quantile transform, we can create the entire pipeline without reference to the data:

pipeline = Select("a", "b", "c") → Quantile()
SequentialTransform
├─ Select(selector: [:a, :b, :c], newnames: nothing)
└─ Quantile(selector: all, dist: Distributions.Normal{Float64}(μ=0.0, σ=1.0))

The operator → creates a special SequentialTransform, which can be applied like any other transform in the framework:

gtb |> pipeline
9990 rows omitted
10000×4 GeoTable over 100×100 CartesianGrid
a b c geometry
Continuous Continuous Continuous Quadrangle
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
0.58017 -0.152998 -0.441571 Quadrangle((x: 0.0 m, y: 0.0 m), ..., (x: 0.0 m, y: 1.0 m))
0.876055 0.174083 1.46912 Quadrangle((x: 1.0 m, y: 0.0 m), ..., (x: 1.0 m, y: 1.0 m))
0.905879 0.100182 0.868721 Quadrangle((x: 2.0 m, y: 0.0 m), ..., (x: 2.0 m, y: 1.0 m))
0.160865 -0.126925 -1.09893 Quadrangle((x: 3.0 m, y: 0.0 m), ..., (x: 3.0 m, y: 1.0 m))
1.92523 -1.74148 -0.20087 Quadrangle((x: 4.0 m, y: 0.0 m), ..., (x: 4.0 m, y: 1.0 m))
1.84526 0.740494 0.0624556 Quadrangle((x: 5.0 m, y: 0.0 m), ..., (x: 5.0 m, y: 1.0 m))
1.21912 -0.503518 0.283492 Quadrangle((x: 6.0 m, y: 0.0 m), ..., (x: 6.0 m, y: 1.0 m))
1.42624 1.95486 -0.949434 Quadrangle((x: 7.0 m, y: 0.0 m), ..., (x: 7.0 m, y: 1.0 m))
0.243007 0.699243 -0.883808 Quadrangle((x: 8.0 m, y: 0.0 m), ..., (x: 8.0 m, y: 1.0 m))
0.639035 -1.59909 0.872749 Quadrangle((x: 9.0 m, y: 0.0 m), ..., (x: 9.0 m, y: 1.0 m))
⋮ ⋮ ⋮ ⋮

It will perform optimizations whenever possible. For instance, we know a priori that adding the Identity transform anywhere in the pipeline doesn’t have any effect:

pipeline → Identity()
SequentialTransform
├─ Select(selector: [:a, :b, :c], newnames: nothing)
└─ Quantile(selector: all, dist: Distributions.Normal{Float64}(μ=0.0, σ=1.0))

7.3 Operator ⊔

The operator ⊔ (\sqcup) can be used to create lazy parallel transforms. There is no equivalent in Julia as this operator is very specific to tables. It combines the geotables produced by two or more pipelines into a single geotable with the disjoint union of all columns.

Let’s illustrate this concept with two pipelines:

pipeline1 = Select("a") → Indicator("a", k=3)
SequentialTransform
├─ Select(selector: [:a], newnames: nothing)
└─ Indicator(selector: :a, k: 3, scale: :quantile, categ: false)
pipeline2 = Select("b", "c", "d") → PCA(maxdim=2)
SequentialTransform
├─ Select(selector: [:b, :c, :d], newnames: nothing)
├─ ZScore(selector: all)
└─ EigenAnalysis(proj: :V, maxdim: 2, pratio: 1.0)

The first pipeline creates 3 indicator variables from variable “a”:

gtb |> pipeline1
9990 rows omitted
10000×4 GeoTable over 100×100 CartesianGrid
a_1 a_2 a_3 geometry
Categorical Categorical Categorical Quadrangle
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
false false true Quadrangle((x: 0.0 m, y: 0.0 m), ..., (x: 0.0 m, y: 1.0 m))
false false true Quadrangle((x: 1.0 m, y: 0.0 m), ..., (x: 1.0 m, y: 1.0 m))
false false true Quadrangle((x: 2.0 m, y: 0.0 m), ..., (x: 2.0 m, y: 1.0 m))
false true true Quadrangle((x: 3.0 m, y: 0.0 m), ..., (x: 3.0 m, y: 1.0 m))
false false true Quadrangle((x: 4.0 m, y: 0.0 m), ..., (x: 4.0 m, y: 1.0 m))
false false true Quadrangle((x: 5.0 m, y: 0.0 m), ..., (x: 5.0 m, y: 1.0 m))
false false true Quadrangle((x: 6.0 m, y: 0.0 m), ..., (x: 6.0 m, y: 1.0 m))
false false true Quadrangle((x: 7.0 m, y: 0.0 m), ..., (x: 7.0 m, y: 1.0 m))
false true true Quadrangle((x: 8.0 m, y: 0.0 m), ..., (x: 8.0 m, y: 1.0 m))
false false true Quadrangle((x: 9.0 m, y: 0.0 m), ..., (x: 9.0 m, y: 1.0 m))
⋮ ⋮ ⋮ ⋮

The second pipeline runs principal component analysis with variables “b”, “c” and “d” and produces 2 principal components:

gtb |> pipeline2
9990 rows omitted
10000×3 GeoTable over 100×100 CartesianGrid
PC1 PC2 geometry
Continuous Continuous Quadrangle
[NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
0.213296 -0.13622 Quadrangle((x: 0.0 m, y: 0.0 m), ..., (x: 0.0 m, y: 1.0 m))
-1.69747 0.153096 Quadrangle((x: 1.0 m, y: 0.0 m), ..., (x: 1.0 m, y: 1.0 m))
-1.13536 0.0866283 Quadrangle((x: 2.0 m, y: 0.0 m), ..., (x: 2.0 m, y: 1.0 m))
1.4932 -0.108655 Quadrangle((x: 3.0 m, y: 0.0 m), ..., (x: 3.0 m, y: 1.0 m))
0.257714 -1.73605 Quadrangle((x: 4.0 m, y: 0.0 m), ..., (x: 4.0 m, y: 1.0 m))
0.0630665 0.695603 Quadrangle((x: 5.0 m, y: 0.0 m), ..., (x: 5.0 m, y: 1.0 m))
-0.453543 -0.473583 Quadrangle((x: 6.0 m, y: 0.0 m), ..., (x: 6.0 m, y: 1.0 m))
1.15646 1.99288 Quadrangle((x: 7.0 m, y: 0.0 m), ..., (x: 7.0 m, y: 1.0 m))
1.69432 0.662209 Quadrangle((x: 8.0 m, y: 0.0 m), ..., (x: 8.0 m, y: 1.0 m))
-1.66776 -1.58832 Quadrangle((x: 9.0 m, y: 0.0 m), ..., (x: 9.0 m, y: 1.0 m))
⋮ ⋮ ⋮

We can combine the two pipelines into a single pipeline that executes in parallel:

pipeline = pipeline1 ⊔ pipeline2
ParallelTableTransform
├─ SequentialTransform
│  ├─ Select(selector: [:a], newnames: nothing)
│  └─ Indicator(selector: :a, k: 3, scale: :quantile, categ: false)
└─ SequentialTransform
   ├─ Select(selector: [:b, :c, :d], newnames: nothing)
   ├─ ZScore(selector: all)
   └─ EigenAnalysis(proj: :V, maxdim: 2, pratio: 1.0)
gtb |> pipeline
9990 rows omitted
10000×6 GeoTable over 100×100 CartesianGrid
a_1 a_2 a_3 PC1 PC2 geometry
Categorical Categorical Categorical Continuous Continuous Quadrangle
[NoUnits] [NoUnits] [NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
false false true 0.213296 -0.13622 Quadrangle((x: 0.0 m, y: 0.0 m), ..., (x: 0.0 m, y: 1.0 m))
false false true -1.69747 0.153096 Quadrangle((x: 1.0 m, y: 0.0 m), ..., (x: 1.0 m, y: 1.0 m))
false false true -1.13536 0.0866283 Quadrangle((x: 2.0 m, y: 0.0 m), ..., (x: 2.0 m, y: 1.0 m))
false true true 1.4932 -0.108655 Quadrangle((x: 3.0 m, y: 0.0 m), ..., (x: 3.0 m, y: 1.0 m))
false false true 0.257714 -1.73605 Quadrangle((x: 4.0 m, y: 0.0 m), ..., (x: 4.0 m, y: 1.0 m))
false false true 0.0630665 0.695603 Quadrangle((x: 5.0 m, y: 0.0 m), ..., (x: 5.0 m, y: 1.0 m))
false false true -0.453543 -0.473583 Quadrangle((x: 6.0 m, y: 0.0 m), ..., (x: 6.0 m, y: 1.0 m))
false false true 1.15646 1.99288 Quadrangle((x: 7.0 m, y: 0.0 m), ..., (x: 7.0 m, y: 1.0 m))
false true true 1.69432 0.662209 Quadrangle((x: 8.0 m, y: 0.0 m), ..., (x: 8.0 m, y: 1.0 m))
false false true -1.66776 -1.58832 Quadrangle((x: 9.0 m, y: 0.0 m), ..., (x: 9.0 m, y: 1.0 m))
⋮ ⋮ ⋮ ⋮ ⋮ ⋮

All 5 columns are present in the final geotable.

7.4 Revertibility

An important concept related to pipelines that is very useful in geospatial data science is revertibility. The concept is useful whenever we need to answer geoscientific questions in terms of variables that have been transformed for geostatistical analysis.

Let’s illustrate the concept with the following geotable and pipeline:

a = [-1.0, 4.0, 1.6, 3.4]
b = [1.6, 3.4, -1.0, 4.0]
c = [3.4, 2.0, 3.6, -1.0]
table = (; a, b, c)

gtb = georef(table, [(0, 0), (1, 0), (1, 1), (0, 1)])
4×4 GeoTable over 4 PointSet
a b c geometry
Continuous Continuous Continuous Point
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
-1.0 1.6 3.4 (x: 0.0 m, y: 0.0 m)
4.0 3.4 2.0 (x: 1.0 m, y: 0.0 m)
1.6 -1.0 3.6 (x: 1.0 m, y: 1.0 m)
3.4 4.0 -1.0 (x: 0.0 m, y: 1.0 m) 
pipeline = Center()
Center transform
└─ selector: all

We saw that our pipelines can be evaluated with Julia’s pipe operator:

gtb |> pipeline
4×4 GeoTable over 4 PointSet
a b c geometry
Continuous Continuous Continuous Point
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
-3.0 -0.4 1.4 (x: 0.0 m, y: 0.0 m)
2.0 1.4 0.0 (x: 1.0 m, y: 0.0 m)
-0.4 -3.0 1.6 (x: 1.0 m, y: 1.0 m)
1.4 2.0 -3.0 (x: 0.0 m, y: 1.0 m)

In order to revert a pipeline, however; we need to save auxiliary constants that were used to transform the data (e.g., mean of selected columns). The apply function serves this purpose:

newgtb, cache = apply(pipeline, gtb)

newgtb
4×4 GeoTable over 4 PointSet
a b c geometry
Continuous Continuous Continuous Point
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
-3.0 -0.4 1.4 (x: 0.0 m, y: 0.0 m)
2.0 1.4 0.0 (x: 1.0 m, y: 0.0 m)
-0.4 -3.0 1.6 (x: 1.0 m, y: 1.0 m)
1.4 2.0 -3.0 (x: 0.0 m, y: 1.0 m)

The function produces the new geotable as usual and an additional cache with all the information needed to revert the transforms in the pipeline. We say that a pipeline isrevertible, if there is an efficient way to revert its transforms starting from any geotable that has the same schema of the geotable produced by the apply function:

isrevertible(pipeline)
true
revert(pipeline, newgtb, cache)
4×4 GeoTable over 4 PointSet
a b c geometry
Continuous Continuous Continuous Point
[NoUnits] [NoUnits] [NoUnits] 🖈 Cartesian{NoDatum}
-1.0 1.6 3.4 (x: 0.0 m, y: 0.0 m)
4.0 3.4 2.0 (x: 1.0 m, y: 0.0 m)
1.6 -1.0 3.6 (x: 1.0 m, y: 1.0 m)
3.4 4.0 -1.0 (x: 0.0 m, y: 1.0 m)

A very common workflow in geospatial data science consists of:

  1. Transforming the data to an appropriate sample space for geostatistical analysis
  2. Doing additional modeling to predict variables in new geospatial locations
  3. Reverting the modeling results with the saved pipeline and cache

We will see examples of this workflow in Part V of the book.

7.5 Congratulations!

Congratulations on finishing Part II of the book. Let’s quickly review what we learned so far:

  • Transforms and pipelines are powerful tools to achieve reproducible geospatial data science.
  • The operators → and ⊔ can be used to build lazy pipelines. After a pipeline is built, it can be applied to different geotables, which may have different types of geospatial domain.
  • Lazy pipelines can always be optimized for computational performance, and the Julia language really thrives to dispatch the appropriate optimizations when they are available.
  • Map projections are specific types of coordinate transforms. They can be combined with many other transforms in the framework to produce advanced geostatistical visualizations.

There is a long journey until the technology reaches its full potential. The good news is that Julia code is easy to read and modify, and you can become an active contributor after just a few weeks working with the language. We invite you to contribute new transforms and optimizations as soon as you feel comfortable with the framework.