Odoo GraphQL Subscription using Node, Express JS for Sample
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 

104 lines
2.7 KiB

import { printBlockString } from '../language/blockString.mjs';
import { isPunctuatorTokenKind, Lexer } from '../language/lexer.mjs';
import { isSource, Source } from '../language/source.mjs';
import { TokenKind } from '../language/tokenKind.mjs';
/**
* Strips characters that are not significant to the validity or execution
* of a GraphQL document:
* - UnicodeBOM
* - WhiteSpace
* - LineTerminator
* - Comment
* - Comma
* - BlockString indentation
*
* Note: It is required to have a delimiter character between neighboring
* non-punctuator tokens and this function always uses single space as delimiter.
*
* It is guaranteed that both input and output documents if parsed would result
* in the exact same AST except for nodes location.
*
* Warning: It is guaranteed that this function will always produce stable results.
* However, it's not guaranteed that it will stay the same between different
* releases due to bugfixes or changes in the GraphQL specification.
*
* Query example:
*
* ```graphql
* query SomeQuery($foo: String!, $bar: String) {
* someField(foo: $foo, bar: $bar) {
* a
* b {
* c
* d
* }
* }
* }
* ```
*
* Becomes:
*
* ```graphql
* query SomeQuery($foo:String!$bar:String){someField(foo:$foo bar:$bar){a b{c d}}}
* ```
*
* SDL example:
*
* ```graphql
* """
* Type description
* """
* type Foo {
* """
* Field description
* """
* bar: String
* }
* ```
*
* Becomes:
*
* ```graphql
* """Type description""" type Foo{"""Field description""" bar:String}
* ```
*/
export function stripIgnoredCharacters(source) {
const sourceObj = isSource(source) ? source : new Source(source);
const body = sourceObj.body;
const lexer = new Lexer(sourceObj);
let strippedBody = '';
let wasLastAddedTokenNonPunctuator = false;
while (lexer.advance().kind !== TokenKind.EOF) {
const currentToken = lexer.token;
const tokenKind = currentToken.kind;
/**
* Every two non-punctuator tokens should have space between them.
* Also prevent case of non-punctuator token following by spread resulting
* in invalid token (e.g. `1...` is invalid Float token).
*/
const isNonPunctuator = !isPunctuatorTokenKind(currentToken.kind);
if (wasLastAddedTokenNonPunctuator) {
if (isNonPunctuator || currentToken.kind === TokenKind.SPREAD) {
strippedBody += ' ';
}
}
const tokenBody = body.slice(currentToken.start, currentToken.end);
if (tokenKind === TokenKind.BLOCK_STRING) {
strippedBody += printBlockString(currentToken.value, {
minimize: true,
});
} else {
strippedBody += tokenBody;
}
wasLastAddedTokenNonPunctuator = isNonPunctuator;
}
return strippedBody;
}