Python xlutils copies over incorrect year

Question

I've written a script that changes values for a certain date in an excel sheet. When I create a new excel file using copy, it copies everything over correctly except for the year portion of the date. For example, it will go from 4/5/2012 to 4/5/2008. It appears that all dates go back four years. Here's the code:

def exceledit():
#open excel sheet                                                           
import xlrd, xlwt, xlutils
import datetime
from xlutils.copy import copy
print 'Opening excel sheet...'
book = xlrd.open_workbook('test.xls', on_demand=True, formatting_info=True)
print 'Creating and editing new excel sheet...'
wbook = copy(book)
print 'Done creating new excel sheet'

sh = book.sheet_by_index(0)
#iterate through dates in excel sheet                                       
for colnum in range(sh.ncols):
    date = sh.cell_value(3, colnum+4)
    #if xlrd finds a date                                                   
    if date:
        #grab date data                                                     
        year, month, day, hour, minute, second =  xlrd.xldate_as_tuple(date\
    , book.datemode)
        #if dates are within the month currently being edited               
        if month == 04:
            #format excel date information to work with parkh dict          
            format =  str(month) + "/" + str(day) + "/" + str(year)        
            print 'Editing ' + format
            #clear cells to eliminate old information                       
            wbook.get_sheet(0).write(6, colnum+6, "")
            wbook.get_sheet(0).write(5, colnum+6, "")
    wbook.get_sheet(0).write(7, colnum+6, "")
            #iterate through hour segments for that day                     
            for x in parkh[format]:
                #if regular hours, insert in "HOURS" row                    
                if x[0] == 'Park Hours':
                    wbook.get_sheet(0).write(6, colnum+6, x[1])
                #if extra magic hours, insert in respective row             
                if x[0] == 'Extra Magic Hours':
                    #insert in morning row                                  
                    if int(x[1][0:1]) in range(2,9):
                        wbook.get_sheet(0).write(5, colnum+6, x[1])
                    #insert in evening row                                  
                    else:
                        wbook.get_sheet(0).write(7, colnum+6, x[1])

        if month == 05:
            break

print 'Done editing. Now saving...'
wbook.save('new.xls')
print 'new.xls saved'

Any idea why it might be changing the year? I haven't seen anyone else have the same issue elsewhere.

Answer 1

You evidently have an input Excel file that is using the 1904 date system. Your immediate problem is that xlutils doesn't support copying these files correctly. Fortunately the fix is a one-liner and you can even do it yourself in your script after you have done the copy:

wbook = copy(book)
wbook.dates_1904 = book.datemode

This works for copying dates because xlwt supports writing the record that specifies what datemode is in use.

WARNING Any new date values that you write into the file by using Worksheet.write() will NOT be written correctly, as xlwt unfortunately ignores the dates_1904 setting when converting datetime.date and datetime.datetime objects to the Excel magic floats.

I have written and tested a fix which replaces the whole body of the Row.__excel_date_dt method. It will be committed to the new xlwt repository on github soonish. In the meantime here is the code if you are in dire need:

def __excel_date_dt(self, date):
    adj = False
    if isinstance(date, dt.date):
        if self.__parent_wb.dates_1904:
            epoch_tuple = (1904, 1, 1)
        else:
            epoch_tuple = (1899, 12, 31)
            adj = True
        if isinstance(date, dt.datetime):
            epoch = dt.datetime(*epoch_tuple)
        else:
            epoch = dt.date(*epoch_tuple)
    else: # it's a datetime.time instance
        epoch = dt.datetime(1900, 1, 1)            
        date = dt.datetime.combine(epoch, date)
    delta = date - epoch
    xldate = delta.days + delta.seconds / 86400.0
    # Add a day for Excel's missing leap day in 1900
    if adj and xldate > 59:
        xldate += 1
    return xldate    

WARNING Trying to convert your file to the 1900 system by opening it in Excel, unticking the 1904 config item and saving the file DOESN'T WORK -- dates will be 4 years out of whack.

What appears to work properly is this:

• open the file in Excel, save it as XML Spreadsheet 2003 (*.xml) ... this format records dates in text format e.g. 1999-12-31T23:59:59.999

• open the XML file in a text editor, find the line that reads
  <Date1904/> ... yes, the XML is human-readable straight out of the box ... and delete it, then save the xml file

• open the changed XML file in Excel, all your data and formatting
  should be preserved, the only difference should be that the pesky
  1904 box is not ticked anymore. You can then save it as an XLS file

Answer 2

I have experienced this with Excel workbooks, not even using Python. When the same file gets passed around between Windows and Mac, weird things can happen with the dates (though this is not typical). Excel for Windows defaults to the so-called "1900" date system, while Excel for Mac defaults to the "1904" date system. (These are the starting years for the respective systems.)

Though the xlrd documentation and source code strongly recommend you stick with the datemode value that is encoded in the workbook (as you have done), I think it's worth a try explicitly giving the "other" mode to xldate_as_tuple to see if it fixes your problem:

year, month, day, hour, minute, second = xlrd.xldate_as_tuple(date,
    1 - book.datemode)

datemode will be 0 for 1900 mode, 1 for 1904 mode; 1 - datemode flips this.

Incidentally, you can break lines without the backslash as long as you are breaking after a comma in a parenthesized expression.