Parsing string values and treating as dates is incorrect

Hi! There is a CSV file, which I try to read and it contains field with value

“Aprobil P 0.1%”

the short example of CSV:

"Country";"Product Family"
"Germany";"Aprobil P 0.1%"

conversion to workbook is the following:

var workbook = XLSX.read(csvData, {
                    type:'string',
                    dateNF: 'D-M-YYYY',
                    cellDates:true,
                    cellText:true,
                    cellNF: false,
                raw:false});

after conversion I save the XLS, where value “Aprobil P 0.1%” is converted to a date 01.04.00

looking into the worksheet model and getting the certain cell, it contains:

{
      t: 'd',
      v: 'Sat Apr 01 2000 00:00:00 GMT+0300 (Eastern European Summer Time)',
      z:undefined
}

is there any way to cover this case? Could you pls assist, what to do.

Thanks in advance!

[SheetJSDev]

Hello @pzhelnov ! First, understand that this is an open source project and all work done on the project is charity. If you are concerned about response time or interested in supporting development, we offer paid builds and support as discussed in https://sheetjs.com/pro

That said, unfortunately there’s no magic solution here. The Date parsing currently falls back on the JS Date builtin features, which are super flexible in chrome. As an example:

new Date("This is not a date 1")

is the date January 1 2001, despite it clearly indicating that it isn’t a date.

Setting the option raw: true will give you the raw strings, suppressing any sort of value interpretation.

If you are interested in helping, the relevant function is fuzzydate. Google Docs / Office Online likely use a series of regular expressions based on some known formats like /\d{4}-\d{2}-\d{2}/ for yyyy-mm-dd and that might be a more sensible approach than using the generic Date constructor

[SheetJSDev]

Based on some testing, in the en-US locale, there appear to be no valid Excel dates that contain letters outside of the month short or long names (even full day names are not accepted). Looking at a number of the issues, a guard like

    var lower = s.toLowerCase();
    if(lower.match(/jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec/)) {
      lower = lower.replace(/[^a-z]/g,"");
      if(lower.length > 3 && lower_months.indexOf(lower) == -1) return n;
      return o;
    }

(where lower_months is an array of the long names of months) would address many of the problems. The proposed “strict” mode is still needed for determining field values and matching formats that Chrome does not accept (for example, 1:23.04:56)