web/views/csvexport.py
author sylvain.thenault@logilab.fr
Wed, 08 Apr 2009 13:17:32 +0200
branchtls-sprint
changeset 1294 870bc725cc9a
parent 824 a5e6acffde30
child 1477 b056a49c16dc
permissions -rw-r--r--
password widget should use the same value as main field for the confirm, else if we don't notice that we got a validation error while no password modification has been done

"""csv export views

:organization: Logilab
:copyright: 2001-2009 LOGILAB S.A. (Paris, FRANCE), all rights reserved.
:contact: http://www.logilab.fr/ -- mailto:contact@logilab.fr
"""
__docformat__ = "restructuredtext en"

from cubicweb.common.uilib import UnicodeCSVWriter
from cubicweb.view import EntityView, AnyRsetView

class CSVMixIn(object):
    """mixin class for CSV views"""
    templatable = False
    content_type = "text/comma-separated-values"    
    binary = True # avoid unicode assertion
    csv_params = {'dialect': 'excel',
                  'quotechar': '"',
                  'delimiter': ';',
                  'lineterminator': '\n'}
    
    def set_request_content_type(self):
        """overriden to set a .csv filename"""
        self.req.set_content_type(self.content_type, filename='cubicwebexport.csv')
            
    def csvwriter(self, **kwargs):
        params = self.csv_params.copy()
        params.update(kwargs)
        return UnicodeCSVWriter(self.w, self.req.encoding, **params)

    
class CSVRsetView(CSVMixIn, AnyRsetView):
    """dumps raw result set in CSV"""
    id = 'csvexport'
    title = _('csv export')
        
    def call(self):
        writer = self.csvwriter()
        writer.writerow(self.columns_labels())
        rset, descr = self.rset, self.rset.description
        eschema = self.schema.eschema
        for rowindex, row in enumerate(rset):
            csvrow = []
            for colindex, val in enumerate(row):
                etype = descr[rowindex][colindex]
                if val is not None and not eschema(etype).is_final():
                    # csvrow.append(val) # val is eid in that case
                    content = self.view('textincontext', rset, 
                                        row=rowindex, col=colindex)
                else:
                    content = self.view('final', rset,
                                        displaytime=True, format='text/plain',
                                        row=rowindex, col=colindex)
                csvrow.append(content)                    
            writer.writerow(csvrow)
    
    
class CSVEntityView(CSVMixIn, EntityView):
    """dumps rset's entities (with full set of attributes) in CSV

    the generated CSV file will have a table per entity type found in the
    resultset. ('table' here only means empty lines separation between table
    contents)
    """
    id = 'ecsvexport'
    title = _('csv entities export')

    def call(self):
        req = self.req
        rows_by_type = {}
        writer = self.csvwriter()
        rowdef_by_type = {}
        for index in xrange(len(self.rset)):
            entity = self.complete_entity(index)
            if entity.e_schema not in rows_by_type:
                rowdef_by_type[entity.e_schema] = [rs for rs, at in entity.e_schema.attribute_definitions()
                                                   if at != 'Bytes']
                rows_by_type[entity.e_schema] = [[display_name(req, rschema.type)
                                                  for rschema in rowdef_by_type[entity.e_schema]]]
            rows = rows_by_type[entity.e_schema]
            rows.append([entity.printable_value(rs.type, format='text/plain')
                         for rs in rowdef_by_type[entity.e_schema]])
        for rows in rows_by_type.itervalues():
            writer.writerows(rows)
            # use two empty lines as separator
            writer.writerows([[], []])