feat(infra): elasticsearch module (#369)

infra/elasticsearch/.eslintignore

@@ -0,0 +1,2 @@
+lib
+node_modules

infra/elasticsearch/.eslintrc.yml

@@ -0,0 +1,3 @@
+---
+extends:
+  - "@socialgouv/eslint-config-recommended"

infra/elasticsearch/.gitignore

@@ -0,0 +1 @@
+node_modules

infra/elasticsearch/lib/analysis.js

@@ -0,0 +1,151 @@
+const { stopwords } = require("./dataset/stop_words");
+const { synonyms } = require("./dataset/synonyms");
+
+const filter = {
+  // Normalize acronyms so that no matter the format, the resulting token will be the same.
+  // E.g.: SmiC => S.M.I.C. => SMIC => smic.
+  french_acronyms: {
+    catenate_all: true,
+    generate_number_parts: false,
+    generate_word_parts: false,
+    type: "word_delimiter",
+  },
+  // Remove elision (l'avion => avion)
+  // ne prend pas en compte la casse (L'avion = l'avion = avion)
+  french_elision: {
+    articles: [
+      "l",
+      "m",
+      "t",
+      "qu",
+      "n",
+      "s",
+      "j",
+      "d",
+      "c",
+      "jusqu",
+      "quoiqu",
+      "lorsqu",
+      "puisqu",
+      "parce qu",
+      "parcequ",
+      "entr",
+      "presqu",
+      "quelqu",
+    ],
+    articles_case: true,
+    type: "elision",
+  },
+
+  // Il existe 3 stemmer pour le francais french, light_french, minimal_french
+  // light french et le median
+  french_stemmer: {
+    language: "light_french",
+    type: "stemmer",
+  },
+
+  french_stop: {
+    stopwords: stopwords,
+    type: "stop",
+  },
+  // liste de termes et leurs synonymes
+  french_synonyms: {
+    expand: true,
+    synonyms: synonyms,
+    type: "synonym",
+  },
+};
+
+const analyzer = {
+  article_id_analyzer: {
+    filter: ["lowercase", "french_acronyms"],
+    tokenizer: "article_id_tokenizer",
+  },
+
+  // used at index time to generate ngrams
+  // for all suggestion
+  // see below, ngram from tokens
+  autocomplete: {
+    filter: ["lowercase", "icu_folding"],
+    tokenizer: "autocomplete", //, "french_stop"]
+  },
+
+  // at search time, we only consider
+  // the entire query (no ngrams)
+  autocomplete_search: {
+    filter: "icu_folding",
+    tokenizer: "lowercase",
+  },
+
+  french: {
+    filter: [
+      "french_elision",
+      "icu_folding",
+      "lowercase",
+      "french_stop",
+      "french_stemmer",
+    ],
+    tokenizer: "icu_tokenizer",
+  },
+
+  french_indexing: {
+    char_filter: ["startwith"],
+    filter: [
+      "french_elision",
+      "icu_folding",
+      "lowercase",
+      "french_stop",
+      "french_stemmer",
+    ],
+    tokenizer: "icu_tokenizer",
+  },
+
+  french_with_synonyms: {
+    char_filter: ["html_strip"],
+    filter: [
+      "french_elision",
+      "icu_folding",
+      "lowercase",
+      "french_synonyms",
+      "french_stop",
+      "french_stemmer",
+    ],
+    tokenizer: "icu_tokenizer",
+  },
+
+  idcc_ape: {
+    tokenizer: "whitespace",
+  },
+  // improve match_phrase_prefix query
+  // using a keyword analyser on type:text field
+  // in order to match results with query as prefix
+  // (as opposite to match "in the middle")
+  sugg_prefix: {
+    char_filter: ["startwith"],
+    filter: ["lowercase", "icu_folding"],
+    tokenizer: "icu_tokenizer",
+  },
+};
+
+const char_filter = {
+  startwith: {
+    pattern: "^(.*)",
+    replacement: "__start__ $1",
+    type: "pattern_replace",
+  },
+};
+
+const tokenizer = {
+  article_id_tokenizer: {
+    pattern: "[0123456789]{4}-[0123456789]{1,3}-?[0123456789]{1,3}?",
+    type: "simple_pattern",
+  },
+  autocomplete: {
+    max_gram: 10,
+    min_gram: 2,
+    token_chars: ["letter"],
+    type: "edge_ngram",
+  },
+};
+
+module.exports = { analyzer, char_filter, filter, tokenizer };

infra/elasticsearch/lib/dataset/stop_words/index.js

@@ -0,0 +1 @@
+exports.stopwords = require("./stop_words.json")