#### Unit testing is painful amirite?

Writing good tests for your code very often means spending twice as much time coding them than on the things you test themselves.

It is good practice though to verify as much as possible that the code you write is valid, especially if that code is going to be public or included in someone else's work.

In my workflow I insist on the notion of ownership :

The bottomline for me is this: if there are several people on a project, I want clearly defined ownership. It's not that I won't fix a bug in someone else's code, just that they own it and therefore have to have a reliable way of testing that my fix works.
Tests solve part of that problem. My code, my tests. If you fix my code, run my tests, I'm fairly confident that you didn't wreck the whole thing. And that I won't have to spend a couple of hours figuring out what it is that you did.

This a a very very very light constraint when you compare it to methodologies like TDD, but it's a required minimum for me.

Plus, it's not that painful, except...

#### Testing every case

In my personal opinion, the tests that are hardest to do right are the ones that have a very large input range, with a few failure/continuity points.

If, for instance, and completely randomly, of course, you had an application where the tilt of the phone changes the state of the app (locked/unlocked, depending on whether the phone is lying flat-ish on the table or not:

• from -20º to 20º the app is locked
• from 160º to 200º the app is locked
• the rest of the time it's not locked
• All of that modulo 360, of course

So you have a function that takes the current pitch angle, and returns if we should lock or not:

func pitchLock(_ angle: Double) -> Bool {
// ...
}

Does it work? Does it work modulo 360? What would a unit test for that function even look like? A for loop?

I have been looking for a way to do that kind of test for a while, which is why I published HoledRange (now Domains 😇) a while back, as part of my hacks.

What I wanted is to write my tests kind of like this (invalid code on so many levels):

for x in [-1000.0...1000.0].randomSelection {
let unitCircleAngle = x%360.0
if unitCircleAngle >= 340 || unitCircle <= 20 {
XCTAssert(pitchLock(x))
} else if unitCircleAngle >= 160 && unitCircle <= 200 {
XCTAssert(pitchLock(x))
} else {
XCTAssertFalse(pitchLock(x))
}
}

This way of testing, while vaguely valid, leaves so many things flaky:

• how many elements in the random selection?
• how can we make certain values untestable (because we address them somewhere else, for instance)
• what a lot of boilerplate if I have multiple functions to test on the same range of values
• I can't reuse the same value for multiple tests to check function chains

#### Function builders

I have been fascinated with @_functionBuilder every since it was announced. While I don't feel enthusiastic about SwiftUI (in french), that way to build elements out of blocks is something I have wanted for years.

Making them is a harrowing experience the first time, but in the end it works!

What I wanted to use as syntax is something like this:

func myPlus(_ a: Int, _ b: Int) -> Int

DomainTests<Int> {
Domain(-10000...10000)
1000000
Test { (a: Int) in
XCTAssert(myPlus(a, 1) == a+1, "Problem with value\(a)")
XCTAssert(myPlus(1, a) == a+1, "Problem with value\(a)")
}
Test { (a: Int) in
let random = Int.random(in: -10000...10000)
XCTAssert(myPlus(a, random) == a+random, "Problem with value\(a)")
XCTAssert(myPlus(random, a) == a+random, "Problem with value\(a)")
}
}.random()

This particular DomainTests runs 1000000 times over $$D=[-10000;10000]$$ in a random fashion.

Note the Test builder that takes a function with a parameter that will be in the domain, and the definition that allows to define both the test domain (mandatory) and the number of random iterations (optional).

If you want to test every single value in a domain, the bounding needs to be Strideable, ie usable in a for-loop.

DomainTests<Int> {
Domain(-10000...10000)
Test { (a: Int) in
XCTAssert(myPlus(a, 1) == a+1, "Problem with value\(a)")
XCTAssert(myPlus(1, a) == a+1, "Problem with value\(a)")
}
Test { (a: Int) in
let random = Int.random(in: -10000...10000)
XCTAssert(myPlus(a, random) == a+random, "Problem with value\(a)")
XCTAssert(myPlus(random, a) == a+random, "Problem with value\(a)")
}
}.full()

#### Conclusion

A couple of hard working days plus a healthy dose of using that framework personally means this should be ready-ish for production.

If you are a maths-oriented dev and shiver at the idea of untested domains, this is for you 😬

I have a weird thing with the multiplication of command-line tools and gizmos: I forget them.

Do I want to run supercool gitlab commands? Hell yea! Do I need to install 12 utilities (or code a new one) to archive every project older than a year? I hope not...

### The setup

I am a sucker for well documented fully linted code. But the thing is, all the gizmos that help me do that have to be installed in the system or in my ~/bin and I have to remember to update them, and I have to install them on my CD machine, and on every new environment I setup, and make sure they are still compatible with the toolchain, and it freaks me out, ok?

Plus,watching the students try to do it is painful.

So, given a 100% vanilla swift-capable environment, can I manage to run documentation and linting?

### The idea

We have Swift Package Manager, which is now a first-class citizen in XCode, but it can't run shell script phases without some nasty hacks.

What if some targets were (wait for it) built to do the documentation and the linting?

### Linting

One of the most popular linters out there is swiftlint, and it supports SPM. It can also build a library instead of an executable, which means one of my targets could just run the linting and output it in the terminal.

In the Package.swift file, all I needed to do was add the right dependency, and the right product and voila!

Now, SPM is very strict with paths, so I had to put a file named main.swift in the Sources/<target>/ directory, in this case Sources/Lint.

Running the linter is fairly straightforward, and goes in the main.swift file:

Setup the .swiftlint file as usual, and run the command via swift run Lint

Sources/WonderfulPackage/main.swift
⛔️ Line 15: Variable name should be between 3 and 40 characters long: 'f'
⚠️ Line 13: Arguments can be omitted when matching enums with associated types if they are not used.
⚠️ Line 12: Line should be 120 characters or less: currently 143 characters

### Documentation

Documentation is actually trickier, because most documentation tools out there aren't built in swift, or compatible with SPM. Doxygen and jazzy are great, but they don't fit my needs.

I found a project that was extremely promising called SourceDocs by Eneko Alonso, but it isn't a library, so I had to fork it and make it into one (while providing a second target to generate the executable if needed). One weird issue is that SPM doesn't like subtargets to bear the same name so I had to rename a couple of them to avoid conflict with Swift Argument Parser (long story).

I finally found myself in the same spot than with the linter. All I needed to do was create another target, and Bob's you're uncle. Well actually he was mine. I digress.

Another well-placed main file:

Now, the command swift run Docs generates the markdown documentation in the Documentation directory.

Parsing main.swift (1/1)
Removing reference documentation at 'WonderfulPackage/Documentation/KituraStarter'... ✔
Generating Markdown documentation...
Writing documentation file: WonderfulPackage/Documentation/WonderfulPackage/structs/WonderfulPackage.md ✔
Done 🎉
Successful run of the documentation phase


### Conclusion

✅ Vanilla swift environment
✅ No install needed
✅ Works on Linux and MacOS
✅ Integrated into SPM
⚠️ When running in XCode, the current directory is always wonky for packages

With 0.8 dropping, a few things in my previous posts changed, thankfully not much. And by trying to train bigger models, I ran into a huge RAM issue, so I'll share what I did in a few paragraphs

##### Changes for 0.8

valueWithGradient is now a global module function, and you have to call it through TensorFlow like this:

let (loss, grad) = TensorFlow.valueWithGradient(at: model) { (model: TextModel) -> Tensor<Float> in
let logits = model(sampleFeatures)
return softmaxCrossEntropy(logits: logits, labels: sampleLabels)
}

Also, they revamped the serialization mechanics, you can now get serializable data through

try model.serializedParameters()
##### RAM issues

It so happens that someone told me to try with characters trigrams instead of word trigrams. I have no idea if the results are better or worst yet, because the dataset generated is huge: 4*<number of chars>, and a somewhat simple text file gave way to a magnificent 96GB of RAM usage.

Of course, this means that the program can't really run. It also meant I had to find an alternative way, and the simplest I know of that I could implement quickly was storing all the trigrams in a big database, and extract random samples from it, rather than doing it in memory. This meants going from 96GB of RAM usage down to 4GB.

##### The setup

I do Kitura stuff, and I ♥️ PostgreSQL, so I went for a simple ORM+Kuery setup.

The table stores trigrams, and I went for generics for the stored structure:

struct StorableTrigram<FScalar, TScalar> : Codable where FScalar : TensorFlowScalar, FScalar : Codable, TScalar : TensorFlowScalar, TScalar : Codable {
var random_id : Int64
var t1 : FScalar
var t2 : FScalar
var t3 : FScalar
var r : TScalar
}

extension StorableTrigram : Model {
static var tableName: String {
get {
return ("StorableTrigram"+String(describing: FScalar.self)+String(describing: TScalar.self)).replacingOccurrences(of: " ", with: "_")
}
}
}

The random_id will be used to shuffle the lines into multiple partitions later, and the tableName override is to avoid < and > from the table name.

##### The partitionning

One of the key things needed to avoid saturating the RAM is to partition the data. As the rest of the training loop expects an array, I decided to go with a custom Collection that can fit in a for loop and load only the current partition:

struct RandomAccessPartition : Collection {
let numberOfPartitions: Int
let db : ConnectionPool

typealias Index = Int
var startIndex: Int { return 0 }
var endIndex: Int { return numberOfPartitions-1 }

func index(after i: Int) -> Int {
return i+1
}

subscript(position: Int) -> (features: Tensor<Float>, labels: Tensor<Int32>) {
let partitionSize = Int64.max / Int64(numberOfPartitions)
let start_rid = partitionSize * Int64(position)
let end_rid = partitionSize * Int64(position + 1)
var rf : [[Float]] = []
var rl : [Int32] = []

let lsem = DispatchSemaphore(value: 0)
db.getConnection() { conn, err in
if conn == nil {
lsem.signal()
return
}

conn!.execute("SELECT * FROM \"\(StorableTrigram<Float,Int32>.tableName)\" WHERE random_id >= \(start_rid) AND random_id < \(end_rid)") { resultSet in
resultSet.asRows { rows,error in
guard let rows = rows else {
lsem.signal()
return
}
for row in rows {
if let t1 = row["t1"] as? Float,
let t2 = row["t1"] as? Float,
let t3 = row["t1"] as? Float,
let r = row["r"] as? Int32 {
rf.append([t1,t2,t3])
rl.append(r)
}
}
lsem.signal()
}
}
}

lsem.wait()
let featuresT = Tensor<Float>(shape: [rf.count, 3], scalars: rf.flatMap { $0 }) let labelsT = Tensor<Int32>(rl) return (featuresT, labelsT) } } Relying on random_id for the partitions is a bit iffy, but thankfully PostgreSQL can re-randomize those ids somewhat fast works well enough for my use ##### The TextBatch replacement The three key features of that batch-holding struct was: • initialization • random sample (once) • random partitions (once every epoch) So here's the relevant code, with breaks for explanations: struct RandomAccessStringStorage { var db : ConnectionPool var tableCreated : Bool = false let original: [String] let vocabulary: [String] let indexHelper: [Int:Int] init(db database: ConnectionPool, original o: [String], terminator: String? = nil, fromScratch: Bool) { db = database Database.default = Database(database) // shady, but hey original = o let f : [[Float]] let l : [Int32] let v : [String] let h : [Int:Int] if let term = terminator { (f,l,v,h) = RandomAccessStringStorage.makeArrays(original, terminator: term) } else { (f,l,v,h) = RandomAccessStringStorage.makeArrays(original) } vocabulary = v indexHelper = h if fromScratch { deleteAll() for i in 0..<f.count { insertTrigram(t1: f[i][0], t2: f[i][1], t3: f[i][2], r: l[i]) } } } mutating func deleteAll() { let _ = try? StorableTrigram<Float,Int32>.dropTableSync() tableCreated = false } mutating func insertTrigram(t1: Float, t2: Float, t3: Float, r: Int32) { if !tableCreated { let _ = try? StorableTrigram<Float,Int32>.createTableSync() tableCreated = true } let trig = StorableTrigram(random_id: Int64.random(in: Int64(0)...Int64.max), t1: t1, t2: t2, t3: t3, r: r) let lsem = DispatchSemaphore(value: 0) trig.save { st, error in lsem.signal() } lsem.wait() } // ... } The two makeArrays are copied and pasted from the in-memory TextBatch, and the only other thing the initialization relies on is the insertion in the DB system. There are two ways of drawing random items: a one-off and partition the data into random chunks: func randomSample(of size: Int) -> (features: Tensor<Float>, labels: Tensor<Int32>) { var rf : [[Float]] = [] var rl : [Int32] = [] let lsem = DispatchSemaphore(value: 0) db.getConnection() { conn, err in if conn == nil { lsem.signal() return } conn!.execute("SELECT * FROM \"\(StorableTrigram<Float,Int32>.tableName)\" ORDER BY random() LIMIT \(size)") { resultSet in resultSet.asRows { rows,error in guard let rows = rows else { lsem.signal() return } for row in rows { if let t1 = row["t1"] as? Float, let t2 = row["t1"] as? Float, let t3 = row["t1"] as? Float, let r = row["r"] as? Int32 { rf.append([t1,t2,t3]) rl.append(r) } } lsem.signal() } } } lsem.wait() let featuresT = Tensor<Float>(shape: [rf.count, 3], scalars: rf.flatMap {$0 })
let labelsT = Tensor<Int32>(rl)
return (featuresT, labelsT)
}

Random selection in Pg actually works pretty well, but can't be repeated, which is why we have to rely on the random_id to partition:

func randomSample(splits: Int) -> RandomAccessPartition<Float,Int32> {
// reshuffle (will take a while)
// update "StorableTrigramFloatInt32" SET random_id = cast(9223372036854775807 * random() as bigint);
let lsem = DispatchSemaphore(value: 0)
db.getConnection() { conn, err in
if conn == nil {
lsem.signal()
return
}

conn!.execute("UPDATE \"\(StorableTrigram<Float,Int32>.tableName)\" SET random_id = cast(9223372036854775807 * random() as bigint)") { resultSet in
lsem.signal()
}
}
lsem.wait()
return RandomAccessPartition<Float,Int32>(numberOfPartitions: splits, db: self.db)
}

The update will re-randomize the ids, paving the way for the RandomAccessPartition.

Of course the tradeoff in terms of performance is rather big, especially in the initialization phase, but hey, more ram to do other things when the model is training!

Every now and again (especially when training a model), I need to have a guesstimate as to how long a "step" takes, and how long the process will take, so I wrote myself a little piece of code that does that. Because I've had the question multiple times (and because I think everyone codes their own after a while), here's mine. Feel free to use it

/// Structure that keeps track of the time it takes to complete steps, to average or estimate the remaining time
public struct TimeRecord {
/// The number of steps to keep for averaging. 5 is a decent default, increase or decrease as needed
/// Minimum for average is 2, obvioulsy
public var smoothing: Int = 5 {
didSet {
smoothing = max(smoothing, 2) // minimum 2 values
}
}
/// dates for the steps
private var dates : [Date] = []
/// formatter for debug print and/or display
private var formatter = DateComponentsFormatter()
public var formatterStyle : DateComponentsFormatter.UnitsStyle {
didSet {
formatter.allowedUnits = [.hour, .minute, .second, .nanosecond]
formatter.unitsStyle = formatterStyle
}
}

public init(smoothing s: Int = 5, style fs: DateComponentsFormatter.UnitsStyle = .positional) {
smoothing = max(s, 2)
formatterStyle = fs
formatter = DateComponentsFormatter()
// not available everywhere
// formatter.allowedUnits = [.hour, .minute, .second, .nanosecond]
formatter.allowedUnits = [.hour, .minute, .second]
formatter.unitsStyle = fs
}

/// adds the record for a step
/// - param d: the date of the step. If unspecified, current date is taken
mutating func addRecord(_ d: Date? = nil) {
if let d = d { dates.append(d) }
else { dates.append(Date()) }
while(dates.count > smoothing) { dates.remove(at: 0) }
}

/// gives the average delta between two steps (in seconds)
var averageDelta : Double {
if dates.count <= 1 { return 0.0 }
var totalTime = 0.0
for i in 1..<dates.count {
totalTime += dates[i].timeIntervalSince(dates[i-1])
}

}

/// gives the average delta between two steps in human readable form
/// - see formatterStyle for options, default is "02:46:40"
let delta = averageDelta
return formatter.string(from: delta) ?? ""
}

/// given a number of remaining steps, gives an estimate of the time left on the process (in s)
func estimatedTimeRemaining(_ steps: Int) -> Double {
return Double(steps) * averageDelta
}

/// given a number of remaining steps, gives an estimate of the time left on the process in human readable form
/// - see formatterStyle for options, default is "02:46:40"
func estimatedTimeRemainingHumanReadable(_ steps: Int) -> String {
let delta = estimatedTimeRemaining(steps)
return formatter.string(from: delta) ?? ""
}
}

When I train a model, I tend to use it that way:

// prepare model
var tt = TimeRecord()

while currentEpoch < maxEpochs {
// train the model
if currentEpoch > 0 && currentEpoch % 5 == 0 {
print(tt.averageDeltaHumanReadable + " per epoch, "
+ tt.(estimatedTimeRemainingHumanReadable(maxEpochs - currentEpoch) + " remaining"
)
}
}

As everyone settles down in the new mode of operations, the number of small tasks has increased and the number of big projects has decreased.

The plagiarism tool is in testing among some of the teachers at school, and the funny reaction of my team of developers asking for an API (to avoid going through the web front end that I crafted - probably badly - in React) made me smile.

What fascinates me overall is the inability of "the web" to cope with the sudden influx of having a ton more people working from home. "They" said the web would replace everything, that it was just a matter of scaling up.

Azure seems to be full, GCloud has some issues with the data traffic, AWS is holding but the status page keep showing outages...

Don't get me wrong, I've been working remote for close to 20 years, so I'm not saying office work is better. But I have been working on projects with people who said it didn't matter if the performance was poor, because they'd just order a bigger server or two.

That inability to take into account the physical constraints of our world is one of the things that grind my gears the most: I've been working on embedded software and high-performance backend stuff for a long time, and betting on poor code hygiene to be compensated by someone else is never a good bet. It ends up with re-writing the code again, and again, and again.

When it's not the RAM issues (lookin at you Electron), it's server constraints (oh the surprise when your instance autoscales up), or bandwidth issues (our government is thinking of restricting the use of Netflix and the like 🙄).

This situation will hopefully remove the attention from the people who can talk and present the best, and back onto more objective metrics (aka "does it work under load?")

I don't rent a small server because I'm cheap. I do it because I can't release any software that doesn't work "correctly" on the bare minimum of specs I have decided the users will have. Then again, when it explodes, it's a great opportunity to learn new things about optimization and constraints 🧐

Since the situation will last a while longer, I hope it reminds everyone that what we do isn't magic. It's science, and we can't wave the problems and constraints away.