Mailspring/internal_packages/thread-search/lib/search-query-parser.es6
Mark Hahnenberg 668c2935c9 [thread-search] Improve local thread search
Summary:
Previously we waited to build the local ThreadSearch index until we were
done with the initial mail sync. This is undesirable because inboxes can
be very large, making local sync useless for a number of hours after a
user adds an account. This diff removes that restriction.

This diff also adds a new rudimentary new grammar (along with accompanying lexer
and parser) for local thread search queries. This grammar is then
translated into the appropriate SQL queries on the ThreadSearch index
table. More advanced features (e.g. in:category, date ranges, etc) can be added
easily in the future by augmenting this simple search query language.

Test Plan: Run locally, new unit tests

Reviewers: juan, evan, khamidou

Reviewed By: khamidou

Differential Revision: https://phab.nylas.com/D3614
2017-01-11 14:26:42 -08:00

299 lines
7 KiB
JavaScript

import {
ThreadQueryAST,
} from 'nylas-exports';
const {
SearchQueryToken,
OrQueryExpression,
AndQueryExpression,
FromQueryExpression,
ToQueryExpression,
SubjectQueryExpression,
GenericQueryExpression,
TextQueryExpression,
UnreadStatusQueryExpression,
StarredStatusQueryExpression,
} = ThreadQueryAST;
const nextStringToken = (text) => {
if (text[0] !== '"') {
throw new Error('Expected string token to begin with double quote (")');
}
if (text.length < 2) {
throw new Error('Expected string but ran out of input');
}
let pos = 1;
while (pos < text.length) {
const c = text[pos];
if (c === '"') {
return [new SearchQueryToken(text.substring(1, pos)), text.substring(pos + 1)];
}
pos += 1;
}
throw new Error('Expected string but ran out of input');
};
const isWhitespace = (c) => {
switch (c) {
case ' ':
case '\t':
case '\n': return true;
default: return false;
}
};
const consumeWhitespace = (text) => {
let pos = 0;
while (pos < text.length && isWhitespace(text[pos])) {
pos += 1;
}
return text.substring(pos);
};
const reserved = [
'(',
')',
':',
'is',
'read',
'unread',
'starred',
'and',
'or',
'from',
'to',
'subject',
];
const mightBeReserved = (text) => {
for (const r of reserved) {
if (r.startsWith(text) || r.toUpperCase().startsWith(text)) {
return true;
}
}
return false;
};
const isValidNonStringChar = (c) => {
switch (c) {
case '(':
case ')':
case ':': return false;
default: return !isWhitespace(c);
}
};
const isValidNonStringText = (text) => {
if (text.length < 1) {
return false;
}
for (const c of text) {
if (!isValidNonStringChar(c)) {
return false;
}
}
return true;
};
const nextToken = (text) => {
const newText = consumeWhitespace(text);
if (newText.length === 0) {
return [null, newText];
}
if (newText[0] === '"') {
return nextStringToken(newText);
}
let isReserved = true;
let pos = 0;
while (pos < newText.length) {
if (isWhitespace(newText[pos])) {
return [new SearchQueryToken(newText.substring(0, pos)), newText.substring(pos)];
}
const curr = newText.substring(0, pos + 1);
if (isReserved) {
// We no longer have a reserved keyword.
if (!mightBeReserved(curr)) {
// We became an invalid non-reserved token so return the previous pos.
if (!isValidNonStringText(curr)) {
return [new SearchQueryToken(newText.substring(0, pos)), newText.substring(pos)];
}
// We're still a valid token but we're no longer reserved.
isReserved = false;
}
} else {
// We're not reserved and we become invalid so go back.
if (!isReserved && !isValidNonStringText(curr)) {
return [new SearchQueryToken(newText.substring(0, pos)), newText.substring(pos)];
}
}
pos += 1;
}
return [new SearchQueryToken(newText.substring(0, pos + 1)), newText.substring(pos + 1)];
};
/*
* query: and_query+
*
* and_query: or_query [and_query_rest]
* and_query_rest: AND and_query
*
* or_query: simple_query [or_query_rest]
* or_query_rest: OR or_query
*
* simple_query: TEXT
* | from_query
* | to_query
* | subject_query
* | paren_query
* | is_query
*
* from_query: FROM COLON TEXT
* to_query: TO COLON TEXT
* subject_query: SUBJECT COLON TEXT
* paren_query: LPAREN query RPAREN
* is_query: IS is_query_rest
* is_query_rest: read_cond
* | starred_cond
* read_cond: READ | UNREAD
* starred_cond: STARRED | UNSTARRED
*
* TEXT: STRING
* | [^\s]+
* STRING: DQUOTE [^"]* DQUOTE
*/
const consumeExpectedToken = (text, token) => {
const [tok, afterTok] = nextToken(text);
if (tok.s !== token) {
throw new Error(`Expected '${token}', got '${tok.s}'`);
}
return afterTok;
};
const parseText = (text) => {
const [tok, afterTok] = nextToken(text);
if (tok === null) {
throw new Error('Expected text but none available');
}
return [new TextQueryExpression(tok), afterTok];
};
const parseIsQuery = (text) => {
const afterColon = consumeExpectedToken(text, ':');
const [tok, afterTok] = nextToken(afterColon);
if (tok === null) {
return null;
}
const tokText = tok.s.toUpperCase();
switch (tokText) {
case 'READ':
case 'UNREAD': {
return [new UnreadStatusQueryExpression(tokText === 'UNREAD'), afterTok];
}
case 'STARRED':
case 'UNSTARRED': {
return [new StarredStatusQueryExpression(tokText === 'STARRED'), afterTok];
}
default: break;
}
return null;
};
let parseQuery = null; // Satisfy our robot overlords.
const parseSimpleQuery = (text) => {
const [tok, afterTok] = nextToken(text);
if (tok.s === '(') {
const [exp, afterExp] = parseQuery(afterTok);
const afterRparen = consumeExpectedToken(afterExp, ')');
return [exp, afterRparen];
}
if (tok.s.toUpperCase() === 'TO') {
const afterColon = consumeExpectedToken(afterTok, ':');
const [txt, afterTxt] = parseText(afterColon);
return [new ToQueryExpression(txt), afterTxt];
}
if (tok.s.toUpperCase() === 'FROM') {
const afterColon = consumeExpectedToken(afterTok, ':');
const [txt, afterTxt] = parseText(afterColon);
return [new FromQueryExpression(txt), afterTxt];
}
if (tok.s.toUpperCase() === 'SUBJECT') {
const afterColon = consumeExpectedToken(afterTok, ':');
const [txt, afterTxt] = parseText(afterColon);
return [new SubjectQueryExpression(txt), afterTxt];
}
if (tok.s.toUpperCase() === 'IS') {
const result = parseIsQuery(afterTok);
if (result !== null) {
return result;
}
}
const [txt, afterTxt] = parseText(text);
return [new GenericQueryExpression(txt), afterTxt];
};
const parseOrQuery = (text) => {
const [lhs, afterLhs] = parseSimpleQuery(text);
const [tok, afterOr] = nextToken(afterLhs);
if (tok === null) {
return [lhs, afterLhs];
}
if (tok.s.toUpperCase() !== 'OR') {
return [lhs, afterLhs];
}
const [rhs, afterRhs] = parseOrQuery(afterOr);
return [new OrQueryExpression(lhs, rhs), afterRhs];
};
const parseAndQuery = (text) => {
const [lhs, afterLhs] = parseOrQuery(text);
const [tok, afterAnd] = nextToken(afterLhs);
if (tok === null) {
return [lhs, afterLhs];
}
if (tok.s.toUpperCase() !== 'AND') {
return [lhs, afterLhs];
}
const [rhs, afterRhs] = parseAndQuery(afterAnd);
return [new AndQueryExpression(lhs, rhs), afterRhs];
};
parseQuery = (text) => {
return parseAndQuery(text);
}
const parseQueryWrapper = (text) => {
let currText = text;
const exps = [];
while (currText.length > 0) {
const [result, leftover] = parseQuery(currText);
exps.push(result);
currText = leftover;
}
if (exps.length === 0) {
throw new Error('Unable to parse query');
}
let result = null;
for (let i = exps.length - 1; i >= 0; --i) {
if (result === null) {
result = exps[i];
} else {
result = new AndQueryExpression(exps[i], result);
}
}
return result;
};
module.exports = {
parseSearchQuery: parseQueryWrapper,
};