mirror of
https://expo.survex.com/repositories/troggle/.git
synced 2024-11-25 16:51:54 +00:00
no need to export cavetab2 anymore
This commit is contained in:
parent
1a62931202
commit
080684e56f
@ -1,49 +0,0 @@
|
||||
import troggle.core.models as models
|
||||
from django.conf import settings
|
||||
|
||||
import csv, re, os
|
||||
|
||||
#format of CAVETAB2.CSV is
|
||||
headers=['KatasterNumber','KatStatusCode','Entrances','UnofficialNumber','MultipleEntrances','AutogenFile','LinkFile','LinkEntrance','Name','UnofficialName',
|
||||
'Comment','Area','Explorers','UndergroundDescription','Equipment','QMList','KatasterStatus','References','UndergroundCentreLine','UndergroundDrawnSurvey',
|
||||
'SurvexFile','Length','Depth','Extent','Notes','EntranceName','TagPoint','OtherPoint','DescriptionOfOtherPoint','ExactEntrance','TypeOfFix','GPSpreSA',
|
||||
'GPSpostSA','Northing','Easting','Altitude','Bearings','Map','Location','Approach','EntranceDescription','PhotoOfLocation','Marking','MarkingComment',
|
||||
'Findability','FindabilityComment']
|
||||
|
||||
def cavetabRow(cave):
|
||||
#mapping of troggle models to table columns is: (guess this could just be a tuple of tuples rather than a dictionary actually)
|
||||
columnsToModelFields={
|
||||
'Name':cave.official_name,
|
||||
'Area':cave.kat_area(),
|
||||
'KatStatusCode':cave.kataster_code,
|
||||
'KatasterNumber':cave.kataster_number,
|
||||
'UnofficialNumber':cave.unofficial_number,
|
||||
#'' : cave.entrances This is a multiple foreignkey now, may be tricky to dump back into csv. Work on this.
|
||||
'Explorers':cave.explorers,
|
||||
'UndergroundDescription':cave.underground_description,
|
||||
'Equipment':cave.equipment,
|
||||
'References':cave.references,
|
||||
'UndergroundDrawnSurvey':cave.survey,
|
||||
'KatasterStatus':cave.kataster_status,
|
||||
'UndergroundCentreLine':cave.underground_centre_line,
|
||||
'Notes':cave.notes,
|
||||
'Length':cave.length,
|
||||
'Depth':cave.depth,
|
||||
'Extent':cave.extent,
|
||||
'SurvexFile':cave.survex_file,
|
||||
}
|
||||
|
||||
caveRow=['' for x in range(len(headers))]
|
||||
for column, modelField in columnsToModelFields.items():
|
||||
if modelField:
|
||||
# Very sorry about the atrocious replace below. I will fix this soon if noone beats me to it. - AC
|
||||
caveRow[headers.index(column)]=modelField.replace(u'\xd7','x').replace(u'\u201c','').replace(u'\u2013','').replace(u'\xbd','')
|
||||
return caveRow
|
||||
|
||||
def writeCaveTab(outfile):
|
||||
cavewriter=csv.writer(outfile,lineterminator='\r')
|
||||
cavewriter.writerow(headers)
|
||||
for cave in models.Cave.objects.all():
|
||||
cavewriter.writerow(cavetabRow(cave))
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user