mirror of
https://github.com/Jermolene/TiddlyWiki5
synced 2024-11-17 23:34:50 +00:00
125 lines
2.7 KiB
JavaScript
125 lines
2.7 KiB
JavaScript
/*\
|
|
title: $:/core/modules/utils/csv.js
|
|
type: application/javascript
|
|
module-type: utils
|
|
|
|
A barebones CSV parser
|
|
|
|
\*/
|
|
(function(){
|
|
|
|
/*jslint node: true, browser: true */
|
|
/*global $tw: false */
|
|
"use strict";
|
|
|
|
var QUOTE = '"';
|
|
|
|
var getCellInfo = function(text, start, length, SEPARATOR) {
|
|
var isCellQuoted = text.charAt(start) === QUOTE;
|
|
var cellStart = isCellQuoted ? start + 1 : start;
|
|
|
|
if (text.charAt(i) === SEPARATOR) {
|
|
return [cellStart, cellStart, false];
|
|
}
|
|
|
|
for (var i = cellStart; i < length; i++) {
|
|
var cellCharacter = text.charAt(i);
|
|
var isEOL = cellCharacter === "\n" || cellCharacter === "\r";
|
|
|
|
if (isEOL && !isCellQuoted) {
|
|
return [cellStart, i, false];
|
|
|
|
} else if (cellCharacter === SEPARATOR && !isCellQuoted) {
|
|
return [cellStart, i, false];
|
|
|
|
} else if (cellCharacter === QUOTE && isCellQuoted) {
|
|
var nextCharacter = i + 1 < length ? text.charAt(i + 1) : '';
|
|
if (nextCharacter !== QUOTE) {
|
|
return [cellStart, i, true];
|
|
} else {
|
|
i++;
|
|
}
|
|
}
|
|
}
|
|
|
|
return [cellStart, i, isCellQuoted];
|
|
}
|
|
|
|
exports.parseCsvString = function(text, options) {
|
|
if (!text) {
|
|
return [];
|
|
}
|
|
|
|
options = options || {};
|
|
var SEPARATOR = options.separator || ",",
|
|
length = text.length,
|
|
rows = [],
|
|
nextRow = [];
|
|
|
|
for (var i = 0; i < length; i++) {
|
|
var cellInfo = getCellInfo(text, i, length, SEPARATOR);
|
|
var cellText = text.substring(cellInfo[0], cellInfo[1]);
|
|
if (cellInfo[2]) {
|
|
cellText = cellText.replace(/""/g, '"');
|
|
cellInfo[1]++;
|
|
}
|
|
nextRow.push(cellText);
|
|
|
|
i = cellInfo[1];
|
|
|
|
var character = text.charAt(i);
|
|
var nextCharacter = i + 1 < length ? text.charAt(i + 1) : '';
|
|
|
|
if (character === "\r" || character === "\n") {
|
|
// Edge case for empty rows
|
|
if (nextRow.length === 1 && nextRow[0] === '') {
|
|
nextRow.length = 0;
|
|
}
|
|
rows.push(nextRow);
|
|
nextRow = [];
|
|
|
|
if (character === "\r") {
|
|
var nextCharacter = i + 1 < length ? text.charAt(i + 1) : '';
|
|
|
|
if (nextCharacter === "\n") {
|
|
i++;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Special case if last cell in last row is an empty cell
|
|
if (text.charAt(length - 1) === SEPARATOR) {
|
|
nextRow.push("");
|
|
}
|
|
|
|
rows.push(nextRow);
|
|
|
|
return rows;
|
|
}
|
|
|
|
/*
|
|
Parse a CSV string with a header row and return an array of hashmaps.
|
|
*/
|
|
exports.parseCsvStringWithHeader = function(text,options) {
|
|
var csv = $tw.utils.parseCsvString(text, options);
|
|
var headers = csv[0];
|
|
|
|
csv = csv.slice(1);
|
|
for (var i = 0; i < csv.length; i++) {
|
|
var row = csv[i];
|
|
var rowObject = Object.create(null);
|
|
|
|
for(var columnIndex=0; columnIndex<headers.length; columnIndex++) {
|
|
var columnName = headers[columnIndex];
|
|
if (columnName) {
|
|
rowObject[columnName] = $tw.utils.trim(row[columnIndex] || "");
|
|
}
|
|
}
|
|
csv[i] = rowObject;
|
|
}
|
|
return csv;
|
|
}
|
|
|
|
})();
|