Daylight savings and getmtime

[Qvx]

Hello,

I'we written a simple web deployment program which scans for the
changes made to local copy of web site. Changed files are than
packaged into a zip file and deployed to web server.

Now here's the catch. Changes are computed using (1) log file from the
last deployment and (2) local file system. Log file contains
datestamps (integers) returned from os.path.getmtim e(f) function at
the time of last deployment. So i'm comparing two getmtime() values.
The problem is when Daylight saving kicks in: suddenly all local files
are reported as older than they were at the time of deployment.

How do I compensate for this?

Thanks,
Tvrtko

For those curious, here is the script. I apologize for Croatian
comments and literals and missing private libraries, but I think the
code is self-explanatory.
______________

# -*- coding: windows-1250 -*-

from os.path import getmtime, join
from os import walk, rename
from zipfile import ZipFile, ZIP_DEFLATED
from sets import Set
from StringIO import StringIO
from ftplib import FTP
from qvx.io import adapt_stdout, unadapt_stdout
from qvx.composite import Dot
from qvx.compositeut il import read_composite
import sys
import time

class DeploymentError (Exception):
pass

class Deployer:
def __init__ (self, cfg_file):
self.reset(cfg_ file)

def reset (self, cfg_file):
self.read_cfg(c fg_file)
self.local_file s = []
self.remote_fil es = []
self.new_files = []
self.deleted_fi les = []
self.newer_file s = []
self.older_file s = []
self.www_all_ti me = None
self.old_deploy ed = False

def read_cfg (self, cfg_file):
tree = read_composite( cfg_file)
self.cfg = Dot(tree).DEPLO YMENT

def prepare_file_li sts (self):
# Sastavi popis _datoteka_ u DIR direktoriju.
# Izostavi datoteke iz _notes direktorija
self.local_file s = []
for root, dirs, files in walk(self.cfg.D IR):
filtered = [join(root, f).replace('\\' , '/') for f in
files if f not in self.cfg.SKIP_F ILES]
self.local_file s.extend(filter ed)
for skip_dir in self.cfg.SKIP_D IRS.split(','):
if skip_dir.strip( ) in dirs:
dirs.remove(ski p_dir)

# Sastavi popis datoteka na serveru
# Koristi se sa informacijama od zadnjeg deploymenta
# Popis se nalazi u www_all.txt datoteci
self.remote_fil es = []
remote_stamps = {}
zip = ZipFile(self.cf g.FILE, 'r')
for line in zip.read('www_a ll.txt').split( '\n'):
name, stamp = line.split('\t' )
remote_stamps[name] = int(stamp)
self.remote_fil es.append(name)
self.www_all_ti me = zip.getinfo('ww w_all.txt').dat e_time

# Deployment nije obavljen ako nije zapisan log
self.old_deploy ed = 'deployment.log ' in zip.namelist()
zip.close()

# Rastavi datoteke u tri kategorije: nove, obrisane i iste
lset = Set(self.local_ files)
rset = Set(self.remote _files)
self.new_files = list(lset - rset)
self.deleted_fi les = list(rset - lset)
common_files = list(lset & rset)

# Pogledaj što se promijenilo u zajedničkim datotekama
self.newer_file s = []
self.older_file s = []
for name in common_files:
remotetime = remote_stamps[name]
localtime = getmtime(name) #+ 3600
# Ako je razlika unutar sekunde, zanemari
if abs(remotetime-localtime) > int(self.cfg.IG NORE_SEC):
if remotetime > localtime:
self.older_file s.append(name)
elif localtime > remotetime:
self.newer_file s.append(name)

def need_redeployme nt (self):
return not self.old_deploy ed

def check_changes (self):
# Ne bi trebalo biti starijih
if self.older_file s:
raise DeploymentError ('Ne smije biti starijih datoteka!')
if not (self.new_files or self.deleted_fi les or
self.newer_file s):
raise DeploymentError ('Nema promjena!')

def make_deployment _file (self):
# Uključi potrebne datoteke
deployment = ZipFile('new_'+ self.cfg.FILE, 'w', ZIP_DEFLATED)
for name in self.new_files + self.newer_file s:
deployment.writ e(name)

# Uključi popis svih datoteka
all_files = '\n'.join([f+'\t'+str(getm time(f)) for f in
self.local_file s])
deployment.writ estr('www_all.t xt', all_files)

# Uključi popis datoteka za obrisati
for_delete = '\n'.join(self. deleted_files)
if for_delete:
deployment.writ estr('www_delet e.txt', for_delete)
deployment.clos e()
print '\nNapravljena je nova deployment datoteka.'

# Preimenuj deployment datoteke
timestr = '%04d-%02d-%02d_%02d-%02d-%02d' % self.www_all_ti me
old_deployment = self.cfg.FILE.r eplace('.zip',
'_'+timestr+'.z ip')
rename(self.cfg .FILE, old_deployment)
rename('new_'+s elf.cfg.FILE, self.cfg.FILE)
print 'Stara deployment datoteka se sada zove', old_deployment

def exec_ftp (self, silent, logtext, func, *arg):
try:
self.ftp_log = self.ftp_log + '\n\n' + logtext
self.ftp_log = self.ftp_log + '\n' + func(*arg)
return 'OK'
except Exception, e:
self.ftp_log = self.ftp_log + '\n' + str(e)
if not silent:
raise DeploymentError (str(e))
return 'ERROR'

def deploy_ftp (self):
self.ftp_log = ''

# Spoji se na FTP server
print '\nSpajam se na ftp server %s ...' % (self.cfg.SERVE R,),
try:
ftp = FTP(self.cfg.SE RVER)
print 'OK'
except Exception, e:
print 'ERROR'
raise DeploymentError ('Ne mogu se spojiti na FTP server:
'+str(e))

# Logiraj se
print 'Logiram se kao %s ... ' % (self.cfg.USER, ),
print self.exec_ftp(F alse, 'LOGIN', ftp.login, self.cfg.USER,
self.cfg.PASSWO RD)

# Kopiraj datoteke
deployment = ZipFile(self.cf g.FILE, 'r')
deployment_file s = [n for n in deployment.name list() if n not
in ['www_all.txt', 'www_delete.txt ', 'deployment.log ']]
if deployment_file s:
print 'Šaljem datoteke:'
for name in deployment_file s:
bytes = deployment.read (name)
fp = StringIO(bytes)
print ' ', name, len(bytes), ' bytes ...',
print self.exec_ftp(T rue, 'STORBIN '+name, ftp.storbinary,
'STOR '+name, fp)
fp.close()

# Obriši datoteke
if 'www_delete.txt ' in deployment.name list():
deleted_files =
deployment.read ('www_delete.tx t').split('\n')
print 'Brišem datoteke:'
for name in deleted_files:
print ' ', name, '...',
print self.exec_ftp(T rue, 'DEL '+name, ftp.delete,
name)
deployment.clos e()

# Bye bye
print 'Završavam s radom ...',
print self.exec_ftp(T rue, 'BYE', ftp.quit)

# Ispiši FTP log
print '\nFTP log:'
print '-'*20,
print self.ftp_log

def write_log (self, text):
# Zapiši deployment log
deployment = ZipFile(self.cf g.FILE, 'a', ZIP_DEFLATED)
deployment.writ estr('deploymen t.log', text)
deployment.clos e()

def run_interactive ly (self):
# Adaptiraj stdout: mijenja kodnu stranicu u cp852 radi ispisa
# hrvatskih grafema, također logira sve poruke
ofa = adapt_stdout('c p852', True)

try:
try:
# Analiziraj datoteke
self.prepare_fi le_lists()
print '*'*5, 'Obrisani', '*'*5, self.deleted_fi les
print '*'*5, 'Novi ', '*'*5, self.new_files
print '*'*5, 'Noviji ', '*'*5, self.newer_file s
print '*'*5, 'Stariji ', '*'*5, self.older_file s

if self.need_redep loyment():
# Pitaj korisnika da li želi poslati od prije
pripremljeni deployment na server
yn = raw_input('\nOd prije pripremljeni deployment
nije obavljen.\n>>> Da li želite taj deployment poslati na server?
[y/N]: ')
if yn.lower().stri p() == 'y':
self.deploy_ftp ()
self.write_log( ''.join(ofa._lo g))
else:
self.check_chan ges()

# Pitaj korisnika da li želi napraviti deployment
datoteku
yn = raw_input('\n>> > Da li želite pripremiti
deployment? [y/N]: ')
if yn.lower().stri p() == 'y':
desc = raw_input('\n>> > Upišite kratki opis:
')
print '\nOpis:', desc

self.make_deplo yment_file()

# Pitaj korisnika da li želi poslati
deployment na server
yn = raw_input('\n>> > Da li želite poslati
deployment na server? [y/N]: ')
if yn.lower().stri p() == 'y':
self.deploy_ftp ()
self.write_log( ''.join(ofa._lo g))
except DeploymentError , e:
print str(e)
finally:
# Vrati stari stdout
unadapt_stdout( ofa)

if __name__ == '__main__':
deployer = Deployer('hpk.i ni')
deployer.run_in teractively()
x = raw_input('\n[Pritisnite Enter ...] ')
sys.exit(0)

[Christophe Cavalaria]

Qvx wrote:

Hello,

I'we written a simple web deployment program which scans for the
changes made to local copy of web site. Changed files are than
packaged into a zip file and deployed to web server.

Now here's the catch. Changes are computed using (1) log file from the
last deployment and (2) local file system. Log file contains
datestamps (integers) returned from os.path.getmtim e(f) function at
the time of last deployment. So i'm comparing two getmtime() values.
The problem is when Daylight saving kicks in: suddenly all local files
are reported as older than they were at the time of deployment.

How do I compensate for this?

Thanks,
Tvrtko

Never use the local time, always use GMT ( UTC ) time instead.

Since it seems os.path.getmtim e already gives UTC time, onemust wonder if
your OS isn't broken in some way ;) If you can't solve that problem, then
use a simple md5sum of the files instead. md5sum isn't that slow to compute
and it gives better results than timestanps.

Or use a specialised tool like rsync which does a very good job for that.

[wittempj]

on my windows xp box os.path.getmtim e gives back local time (I just
saved the file):

os.path.getmtim e('c:\\temp\\te stset.py') 1106955016 print time.mktime(tim e.localtime()) 1106955034.0

You can try to figure out if DST is on by comparing time.localtime( )
versus time.gmtime(). In the Western European Timezone there's one hour
difference in winter and two hours in summer.
os.path.getmtim e('c:\\temp\\te stset.py') 1106954702 g = time.mktime(tim e.gmtime())
l = time.mktime(tim e.localtime())
print g, l 1106951381.0 1106954987.0 os.path.getmtim e('c:\\temp\\te stset.py') 1106955016 print time.mktime(tim e.localtime()) 1106955034.0 print l - g 3606.0

[Nick Coghlan]

wi******@hotmai l.com wrote:

on my windows xp box os.path.getmtim e gives back local time (I just
saved the file):

os.path.get mtime('c:\\temp \\testset.py')
1106955016
print time.mktime(tim e.localtime())

1106955034.0

No. mktime is converting the local time to UTC. Try this instead:

Py> time.gmtime(os. path.getmtime(" c:/devel/mtime_test.txt" ))
(2005, 1, 29, 3, 35, 37, 5, 29, 0)

Py> time.localtime( os.path.getmtim e("c:/devel/mtime_test.txt" ))
(2005, 1, 29, 15, 35, 37, 5, 29, 1)

(This is with my machine set to Sydney time, so that I'm at UTC + 10, with
daylight savings currently active)
You can try to figure out if DST is on by comparing time.localtime( )
versus time.gmtime(). In the Western European Timezone there's one hour
difference in winter and two hours in summer.

To figure out if DST is currently active, you can use:

Py> import time
Py> time.localtime( )
(2005, 1, 29, 14, 29, 26, 5, 29, 0)
Py> time.localtime( ).tm_isdst
0

Switching to Sydney time (which actually uses DST, unlike Brisbane) gives:

Py> import time
Py> time.localtime( ).tm_isdst
1

The same trick will work on historical dates, too. For instance, consider a file
created last winter, when Sydney was NOT on daylight savings:

Py> time.localtime( os.path.getmtim e("c:/log.txt"))
(2004, 7, 20, 19, 46, 35, 1, 202, 0)

Cheers,
Nick.

--
Nick Coghlan | nc******@email. com | Brisbane, Australia
---------------------------------------------------------------
http://boredomandlaziness.skystorm.net