DateKeys/scripts/recovery/words_test.go

package main

import (
	"encoding/hex"
	"io"
	"os"
	"path/filepath"
	"slices"
	"strings"
	"testing"
)

// unicodeDataPath is the copy of UnicodeData.txt of Unicode 18.0.0 that
// internal/pathrule/gen reads; it is not committed. The tests of the full
// normalization skip without it.
const unicodeDataPath = "../../.cache/unicode/18.0.0/UnicodeData.txt"

func loadUnicodeData(t *testing.T) *unicodeData {
	t.Helper()
	b, err := os.ReadFile(unicodeDataPath)
	if err != nil {
		t.Skipf("no UnicodeData.txt of Unicode 18.0.0 in %s: %v", unicodeDataPath, err)
	}
	u, err := parseUnicodeData(b)
	if err != nil {
		t.Fatal(err)
	}
	return u
}

// The vectors of the annex, 79.7: the words without tables and their key,
// for the chain hash of Quicknet, round 1000 and capsule_id 00 to 0f.
func TestAnnexWordVectors(t *testing.T) {
	id := unhex(t, "000102030405060708090a0b0c0d0e0f")
	for _, c := range []struct{ text, words, key string }{
		{"perro luna casa verde tren mar", "perro luna casa verde tren mar", "fceec4d8ca8de86c85a1f26ed49f82a2b38431bd0ce36db995ae7dfd49b96e41"},
		{"\u00d1and\u00fa  PING\u00dcINO\tcami\u00f3n \u00e1rbol \u00c9ter ola", "nandu pinguino camion arbol eter ola", "273295d29370126a3be50b743132718d3cd9137fb3bb4cb20aa23163d2e19bb7"},
		{"N\u0303andu\u0301  PINGU\u0308INO\tcamio\u0301n a\u0301rbol E\u0301ter ola", "nandu pinguino camion arbol eter ola", "273295d29370126a3be50b743132718d3cd9137fb3bb4cb20aa23163d2e19bb7"},
	} {
		words, ok := normalizeSimple(c.text)
		if !ok || strings.Join(words, " ") != c.words {
			t.Fatalf("%q: %q, %v", c.text, words, ok)
		}
		key, err := wordKey(words, 1000, id)
		if err != nil || hex.EncodeToString(key) != c.key {
			t.Fatalf("%q: key %x, %v", c.text, key, err)
		}
	}
}

// The recipe without tables refuses what it does not cover, and then the
// words need UnicodeData.txt.
func TestNormalizeSimpleRefuses(t *testing.T) {
	for _, text := range []string{"\u00e7a va", "stra\u00dfe", "\u00e0 la", "\u03c3\u03b1\u03c3", "a\u00a0b", "\u212b", "a\u200bb"} {
		if _, ok := normalizeSimple(text); ok {
			t.Errorf("%q: the recipe without tables does not cover it", text)
		}
		if _, err := normalizeWords(text, nil); err == nil || !strings.Contains(err.Error(), "-unicodedata") {
			t.Errorf("%q: %v", text, err)
		}
	}
	// The whole of printable ASCII stays, and only A to Z change.
	var ascii []rune
	for r := rune(0x21); r <= 0x7e; r++ {
		ascii = append(ascii, r)
	}
	words, ok := normalizeSimple(string(ascii))
	if !ok || len(words) != 1 || words[0] != strings.ToLower(string(ascii)) {
		t.Errorf("printable ASCII: %q, %v", words, ok)
	}
}

// The full normalization gives the words of every case of wordkey.json, and
// the recipe without tables, where it applies, the same.
func TestNormalizeVectors(t *testing.T) {
	u := loadUnicodeData(t)
	var f struct {
		Normalize []struct {
			Name  string   `json:"name"`
			Text  string   `json:"text"`
			Words []string `json:"words"`
		} `json:"normalize"`
	}
	readJSON(t, filepath.Join(vectorsDir, "wordkey.json"), &f)
	if len(f.Normalize) < 40 {
		t.Fatalf("%d cases of normalize", len(f.Normalize))
	}
	simple := 0
	for _, c := range f.Normalize {
		got := u.normalize(c.Text)
		if got == nil {
			got = []string{}
		}
		if !slices.Equal(got, c.Words) {
			t.Errorf("%s: %q, want %q", c.Name, got, c.Words)
		}
		if words, ok := normalizeSimple(c.Text); ok {
			simple++
			if words == nil {
				words = []string{}
			}
			if !slices.Equal(words, c.Words) {
				t.Errorf("%s, without tables: %q, want %q", c.Name, words, c.Words)
			}
		}
	}
	if simple < 8 {
		t.Errorf("only %d cases without tables", simple)
	}
}

// Only UnicodeData.txt of Unicode 18.0.0 is accepted: another version can
// give other words.
func TestUnicodeDataVersion(t *testing.T) {
	b, err := os.ReadFile(unicodeDataPath)
	if err != nil {
		t.Skip(err)
	}
	b[len(b)-2] ^= 1
	if _, err := parseUnicodeData(b); err == nil || !strings.Contains(err.Error(), "not that of Unicode 18.0.0") {
		t.Fatalf("a changed UnicodeData.txt: %v", err)
	}
}

// format3_time_and_key_words opens with the text of its words, which needs
// no tables, and with UnicodeData.txt too.
func TestRecoverWithWords(t *testing.T) {
	rec, dkc, relObj, _ := loadFixture(t, "format3_time_and_key_words")
	if rec.WordsText == "" {
		t.Fatal("the record has no words_text")
	}
	text := rec.WordsText
	res, err := recoverCapsule(dkc, relObj, credential{words: &text}, io.Discard)
	if err != nil || res.format != 3 || len(res.body.files) != 1 {
		t.Fatalf("%v", err)
	}
	other := "nandu pinguino camion arbol eter mar"
	if _, err := recoverCapsule(dkc, relObj, credential{words: &other}, io.Discard); err == nil || !strings.Contains(err.Error(), "access layer") {
		t.Fatalf("other words: %v", err)
	}
	if b, err := os.ReadFile(unicodeDataPath); err == nil {
		u, err := parseUnicodeData(b)
		if err != nil {
			t.Fatal(err)
		}
		if _, err := recoverCapsule(dkc, relObj, credential{words: &text, ucd: u}, io.Discard); err != nil {
			t.Fatal(err)
		}
	}
}

type ioDiscard struct{}

func (ioDiscard) Write(p []byte) (int, error) { return len(p), nil }