Further improved search matches

This commit is contained in:
Tom Moor
2022-11-27 10:46:54 -05:00
parent fa8685d241
commit d22b44dcff
4 changed files with 92 additions and 29 deletions

View File

@@ -150,6 +150,7 @@
"passport-slack-oauth2": "^1.1.1",
"pg": "^8.5.1",
"pg-hstore": "^2.3.4",
"pg-tsquery": "^8.4.0",
"polished": "^3.7.2",
"prosemirror-commands": "1.2.2",
"prosemirror-dropcursor": "^1.4.0",

View File

@@ -1,6 +1,7 @@
import removeMarkdown from "@tommoor/remove-markdown";
import invariant from "invariant";
import { find, map } from "lodash";
import queryParser from "pg-tsquery";
import { Op, QueryTypes } from "sequelize";
import { DateFilter } from "@shared/types";
import unescape from "@shared/utils/unescape";
@@ -54,12 +55,16 @@ type Results = {
};
export default class SearchHelper {
/**
* The maximum length of a search query.
*/
public static maxQueryLength = 1000;
public static async searchForTeam(
team: Team,
query: string,
options: SearchOptions = {}
): Promise<SearchResponse> {
const wildcardQuery = `${this.escapeQuery(query)}:*`;
const {
snippetMinWords = 20,
snippetMaxWords = 30,
@@ -103,7 +108,7 @@ export default class SearchHelper {
// Build the SQL query to get result documentIds, ranking, and search term context
const whereClause = `
"searchVector" @@ websearch_to_tsquery('english', :query) AND
"searchVector" @@ to_tsquery('english', :query) AND
"teamId" = :teamId AND
"collectionId" IN(:collectionIds) AND
${documentClause}
@@ -113,8 +118,8 @@ export default class SearchHelper {
const selectSql = `
SELECT
id,
ts_rank(documents."searchVector", websearch_to_tsquery('english', :query)) as "searchRanking",
ts_headline('english', "text", websearch_to_tsquery('english', :query), :headlineOptions) as "searchContext"
ts_rank(documents."searchVector", to_tsquery('english', :query)) as "searchRanking",
ts_headline('english', "text", to_tsquery('english', :query), :headlineOptions) as "searchContext"
FROM documents
WHERE ${whereClause}
ORDER BY
@@ -130,7 +135,7 @@ export default class SearchHelper {
`;
const queryReplacements = {
teamId: team.id,
query: wildcardQuery,
query: this.webSearchQuery(query),
collectionIds,
documentIds,
headlineOptions: `MaxFragments=1, MinWords=${snippetMinWords}, MaxWords=${snippetMaxWords}`,
@@ -176,8 +181,6 @@ export default class SearchHelper {
limit = 15,
offset = 0,
} = options;
const wildcardQuery = `${SearchHelper.escapeQuery(query)}:*`;
// Ensure we're filtering by the users accessible collections. If
// collectionId is passed as an option it is assumed that the authorization
// has already been done in the router
@@ -197,7 +200,7 @@ export default class SearchHelper {
// Build the SQL query to get documentIds, ranking, and search term context
const whereClause = `
"searchVector" @@ websearch_to_tsquery('english', :query) AND
"searchVector" @@ to_tsquery('english', :query) AND
"teamId" = :teamId AND
${
collectionIds.length
@@ -226,8 +229,8 @@ export default class SearchHelper {
const selectSql = `
SELECT
id,
ts_rank(documents."searchVector", websearch_to_tsquery('english', :query)) as "searchRanking",
ts_headline('english', "text", websearch_to_tsquery('english', :query), :headlineOptions) as "searchContext"
ts_rank(documents."searchVector", to_tsquery('english', :query)) as "searchRanking",
ts_headline('english', "text", to_tsquery('english', :query), :headlineOptions) as "searchContext"
FROM documents
WHERE ${whereClause}
ORDER BY
@@ -245,7 +248,7 @@ export default class SearchHelper {
teamId: user.teamId,
userId: user.id,
collaboratorIds: options.collaboratorIds,
query: wildcardQuery,
query: this.webSearchQuery(query),
collectionIds,
dateFilter,
headlineOptions: `MaxFragments=1, MinWords=${snippetMinWords}, MaxWords=${snippetMaxWords}`,
@@ -302,9 +305,29 @@ export default class SearchHelper {
};
}
/**
* Convert a user search query into a format that can be used by Postgres
*
* @param query The user search query
* @returns The query formatted for Postgres ts_query
*/
private static webSearchQuery(query: string): string {
// limit length of search queries as we're using regex against untrusted input
const limitedQuery = this.escapeQuery(query.slice(0, this.maxQueryLength));
// if the search term is one unquoted word then allow partial matches automatically
const queryWordCount = limitedQuery.split(" ").length;
const singleUnquotedSearch =
queryWordCount === 1 && !limitedQuery.startsWith('"');
return queryParser({ singleQuoteReplacement: "&" })(
singleUnquotedSearch ? `${limitedQuery}*` : limitedQuery
);
}
private static escapeQuery(query: string): string {
// replace "\" with escaped "\\" because sequelize.escape doesn't do it
// https://github.com/sequelize/sequelize/issues/2950
return sequelize.escape(query).replace(/\\/g, "\\\\");
return query.replace(/\\/g, "\\\\");
}
}

View File

@@ -1243,7 +1243,7 @@ describe("#documents.search", () => {
const res = await server.post("/api/documents.search", {
body: {
token: user.getJwtToken(),
query: "sear &",
query: "sear",
},
});
const body = await res.json();
@@ -1254,24 +1254,58 @@ describe("#documents.search", () => {
expect(body.data[2].document.id).toEqual(thirdResult.id);
});
it("should strip junk from search term", async () => {
const user = await buildUser();
const firstResult = await buildDocument({
title: "search term",
text: "this is some random text of the document body",
userId: user.id,
teamId: user.teamId,
describe("search operators", () => {
it("negative search operator", async () => {
const { user } = await seed();
await buildDocument({
title: "search term",
text: "random text",
userId: user.id,
teamId: user.teamId,
});
const firstResult = await buildDocument({
title: "title text",
text: "search term",
userId: user.id,
teamId: user.teamId,
});
const res = await server.post("/api/documents.search", {
body: {
token: user.getJwtToken(),
query: `search -random`,
},
});
const body = await res.json();
expect(res.status).toEqual(200);
expect(body.data.length).toEqual(1);
expect(body.data[0].document.id).toEqual(firstResult.id);
});
const res = await server.post("/api/documents.search", {
body: {
token: user.getJwtToken(),
query: "rando &\\;:()",
},
it("quoted search operator", async () => {
const { user } = await seed();
await buildDocument({
title: "document one",
text: "term search",
userId: user.id,
teamId: user.teamId,
});
const firstResult = await buildDocument({
title: "search term",
text: "content",
userId: user.id,
teamId: user.teamId,
});
const res = await server.post("/api/documents.search", {
body: {
token: user.getJwtToken(),
query: `"search term"`,
},
});
const body = await res.json();
expect(res.status).toEqual(200);
expect(body.data.length).toEqual(1);
expect(body.data[0].document.id).toEqual(firstResult.id);
});
const body = await res.json();
expect(res.status).toEqual(200);
expect(body.data.length).toEqual(1);
expect(body.data[0].document.id).toEqual(firstResult.id);
});
it("should not return draft documents", async () => {

View File

@@ -12014,6 +12014,11 @@ pg-protocol@^1.4.0:
resolved "https://registry.yarnpkg.com/pg-protocol/-/pg-protocol-1.4.0.tgz#43a71a92f6fe3ac559952555aa3335c8cb4908be"
integrity sha512-El+aXWcwG/8wuFICMQjM5ZSAm6OWiJicFdNYo+VY3QP+8vI4SvLIWVe51PppTzMhikUJR+PsyIFKqfdXPz/yxA==
pg-tsquery@^8.4.0:
version "8.4.0"
resolved "https://registry.yarnpkg.com/pg-tsquery/-/pg-tsquery-8.4.0.tgz#411293cce23ca1eeb8c29109af9fadf28f20a7d9"
integrity sha512-m0jIxUVwLKSdmOAlqtlbo6K+EFIOZ/hyOMnoe8DmYFqEmOmvafIjGQFmcPP+z5MWd/p7ExxoKNIL31gmM+CwxQ==
pg-types@^2.1.0:
version "2.2.0"
resolved "https://registry.yarnpkg.com/pg-types/-/pg-types-2.2.0.tgz#2d0250d636454f7cfa3b6ae0382fdfa8063254a3"