From 4b1eb1d64f732ef305391acc5ce0c253f8ef1919 Mon Sep 17 00:00:00 2001
From: Vincent Weevers <mail@vincentweevers.nl>
Date: Sat, 8 Jun 2019 22:35:31 +0200
Subject: [PATCH] Add self-distribution test

---
 README.md                            | 18 +++++++
 benchmarks/index.js                  |  1 +
 benchmarks/self-distribution.js      | 49 ++++++++++++++++++
 benchmarks/self-distribution.plot.js | 76 ++++++++++++++++++++++++++++
 level-bench.js                       |  6 ++-
 lib/result.js                        |  3 +-
 package.json                         |  1 +
 7 files changed, 151 insertions(+), 3 deletions(-)
 create mode 100644 benchmarks/self-distribution.js
 create mode 100644 benchmarks/self-distribution.plot.js

diff --git a/README.md b/README.md
index 4f1269c..c3fdfec 100644
--- a/README.md
+++ b/README.md
@@ -151,6 +151,24 @@ Same as `write`, but in batches rather than singular puts. Perform concurrent `b
 - `--concurrency`: default 1
 - `--valueSize`: size of value, as a number in bytes or string with unit (e.g. `--valueSize 1kb`)
 
+### `self-distribution`
+
+_Not a benchmark, but a temporary cheat to reuse the tooling we have here to test (and visualize) some of the internals. Needs a valid `target` argument, same as real benchmarks, although that argument is not actually used._
+
+Generate keys with a certain order and probability distribution. Options:
+
+- `-n`: amount of keys to generate, default 5e3
+- Other options are passed to [`keyspace`](https://github.com/vweevers/keyspace)
+
+Example:
+
+```
+level-bench run self-distribution memdown -b [--distribution zipfian --skew 1]
+level-bench run self-distribution memdown -b [--distribution zipfian --skew=-1]
+level-bench run self-distribution memdown -b [--keys seq]
+level-bench plot self-distribution
+```
+
 <!-- ### Other ideas
 
 - Write batches in different sizes (feature: define a matrix)
diff --git a/benchmarks/index.js b/benchmarks/index.js
index f8248c5..02ff470 100644
--- a/benchmarks/index.js
+++ b/benchmarks/index.js
@@ -5,3 +5,4 @@ exports['write-random'] = require('./write-random')
 exports['write-sorted'] = require('./write-sorted')
 
 exports['batch-put'] = require('./batch-put')
+exports['self-distribution'] = require('./self-distribution')
diff --git a/benchmarks/self-distribution.js b/benchmarks/self-distribution.js
new file mode 100644
index 0000000..f1bd78b
--- /dev/null
+++ b/benchmarks/self-distribution.js
@@ -0,0 +1,49 @@
+'use strict'
+
+const keyspace = require('keyspace')
+
+exports.defaults = {
+  benchmark: {
+    n: 5e3
+  }
+}
+
+exports.plot = require('./self-distribution.plot')
+
+exports.run = function (factory, stream, options) {
+  stream.write('Step, Key, Frequency\n')
+
+  const generator = keyspace(options.n, Object.assign({}, options, {
+    keyAsNumber: true
+  }))
+
+  let step = 0
+
+  const frequencies = new Array(options.n).fill(0)
+  const keys = new Array(options.n).fill(0)
+
+  while (step < options.n) {
+    const key = generator.key(step)
+
+    frequencies[key]++
+    keys[step++] = key
+
+    if (step % 1000 === 0) {
+      console.log('%d %d%', step, Math.round(step / options.n * 100))
+    }
+  }
+
+  step = 0
+  write()
+
+  function write () {
+    if (step < options.n) {
+      const key = keys[step]
+      const frequency = frequencies[step]
+
+      stream.write(`${step++}, ${key}, ${frequency}\n`, write)
+    } else {
+      stream.end()
+    }
+  }
+}
diff --git a/benchmarks/self-distribution.plot.js b/benchmarks/self-distribution.plot.js
new file mode 100644
index 0000000..0628bcd
--- /dev/null
+++ b/benchmarks/self-distribution.plot.js
@@ -0,0 +1,76 @@
+'use strict'
+
+const e = require('../lib/escape-gnuplot-string')
+
+module.exports = function (title, description, results) {
+  const minHeight = 800
+  const height = minHeight + results.length * 300
+  const subtitleY = 1 - (80 / height)
+  const mainPlotRatio = minHeight / height
+  const freqPlotsRatio = 1 - mainPlotRatio
+  const freqPlotsbHeight = freqPlotsRatio / results.length
+
+  const keys = results.map(function (res, i) {
+    const file = res.csvFile
+    const title = results.length === 1 ? '' : res.id(results, 'default')
+
+    return `'${e(file)}' using ($1):($2) title '${e(title)}' ls ${i + 1} axes x1y1`
+  })
+
+  const frequencies = results.map(function (res, i) {
+    // Sequential keys all have a frequency of 1 which gnuplot can't handle
+    if (/seq/.test(res.meta.options.benchmark.keys)) return
+
+    const file = res.csvFile
+    const title = results.length === 1 ? '' : res.id(results, 'default')
+
+    return [
+      `set size 1,${freqPlotsbHeight.toFixed(3)}`,
+      `set origin 0.0,${((results.length - i - 1) * freqPlotsbHeight).toFixed(3)}`,
+      `plot '${e(file)}' using 1:3 with boxes title '' ls ${i + 1} axes x1y1`
+    ].join('\n    ')
+  }).filter(Boolean)
+
+  return `
+  reset
+  set terminal pngcairo truecolor enhanced font "Ubuntu Mono,10" size 1920,${height} background rgb "#1b1b1b"
+  set datafile separator ','
+
+  set autoscale y
+  set ytics mirror
+  set tics in
+  set xlabel "Step" tc rgb "#999999"
+
+  set key outside tc rgb "#999999"
+  set border lc rgb "#999999"
+
+  # To plot more than 5 files, add more line styles
+  set style line 1 lt 7 ps 0.8 lc rgb "#00FFFF"
+  set style line 2 lt 7 ps 0.8 lc rgb "#D84797"
+  set style line 3 lt 7 ps 0.8 lc rgb "#23CE6B"
+  set style line 4 lt 7 ps 0.8 lc rgb "#F5B700"
+  set style line 5 lt 7 ps 0.8 lc rgb "#731DD8"
+
+  set multiplot
+    set lmargin at screen 0.1
+
+    set size 1,${mainPlotRatio.toFixed(3)}
+    set origin 0,${freqPlotsRatio.toFixed(3)}
+    set title '${e(title)}' tc rgb "#cccccc" offset 0,0.7 font "Ubuntu Mono,12"
+    set label 1 '${e(description)}' tc rgb "#999999" at screen 0.5,${subtitleY.toFixed(3)} center front
+    set ylabel 'Key' tc rgb "#999999"
+    set nologscale y
+    plot ${keys.join(', ')}
+
+    set title ''
+    set label 1 ''
+    set rmargin at screen 0.97
+    set ylabel 'Frequency' tc rgb "#999999"
+    set xlabel 'Key' tc rgb "#999999"
+    set logscale y
+    set xrange [0:*]
+    set style fill solid
+
+    ${frequencies.join('\n\n    ')}
+  unset multiplot`
+}
diff --git a/level-bench.js b/level-bench.js
index 1e41113..b3ae6e4 100644
--- a/level-bench.js
+++ b/level-bench.js
@@ -47,9 +47,11 @@ if (command === 'run') {
     process.exit()
   }
 
+  const type = /^self-/.test(benchmark) ? 'test' : 'benchmark'
+  const mainProps = type === 'test' ? ['harness'] : ['context', 'platform']
   const results = files.map(Result.fromFile)
-  const group = results[0].group(results, { include: ['context', 'platform'] })
-  const title = [`benchmark ${benchmark}`, group].filter(Boolean).join(' on ')
+  const group = results[0].group(results, { include: mainProps })
+  const title = [`${type} ${benchmark}`, group].filter(Boolean).join(' on ')
   const desc = results[0].group(results, { exclude: ['context', 'platform', 'harness'] }) || ''
   const pngFile = path.resolve('.', argv.out || `.benchmarks/${benchmark}.${Date.now()}.png`)
   const plt = benchmarks[benchmark].plot(title, desc, results)
diff --git a/lib/result.js b/lib/result.js
index 487eda0..f33ed58 100644
--- a/lib/result.js
+++ b/lib/result.js
@@ -15,9 +15,10 @@ class Result {
     return new Result(csvFile, meta)
   }
 
-  id (benches) {
+  id (benches, fallback) {
     return this.meta.name ||
       this.meta.id(benches.map(b => b.meta)) ||
+      fallback ||
       'baseline'
   }
 
diff --git a/package.json b/package.json
index 81e19d0..4dee35b 100644
--- a/package.json
+++ b/package.json
@@ -20,6 +20,7 @@
     "fast-deep-equal": "~2.0.1",
     "fast-glob": "~2.2.7",
     "human-number": "~1.0.3",
+    "keyspace": "^0.0.1",
     "level-test": "~7.0.0",
     "mkdirp": "~0.5.1",
     "monotonic-timestamp": "0.0.9",