Get min and max dates

[DFS]

dts= ['10-Mar-1998',
'20-Aug-1997',
'06-Sep-2009',
'23-Jan-2010',
'12-Feb-2010',
'05-Nov-2010',
'03-Sep-2009',
'07-Nov-2014',
'08-Mar-2013']

Of course, the naive:
min(dates) = '03-Sep-2009'
max(dates) = '23-Jan-2010'
is wrong.

Not wanting to use any date parsing libraries, I came up with:
========================================================================
m=[('Dec','12'),('Nov','11'),('Oct','10'),('Sep','09'),
('Aug','08'),('Jul','07'),('Jun','06'),('May','05'),
('Apr','04'),('Mar','03'),('Feb','02'),('Jan','01')]

#create new list with properly sortable date (YYYYMMDD)
dts2 = []
for d in dts:
dts2.append((d[-4:]+dict(m)[d[3:6]]+d[:2],d))

print 'min: ' + min(dts2)[1]
print 'max: ' + max(dts2)[1]
========================================================================
$python getminmax.py
min: 20-Aug-1997
max: 07-Nov-2014

which is correct, but I sense a more pythonic way, or a one-liner list
comprehension, is in there somewhere.

Any ideas? Thanks

[Steven D'Aprano]

On Thursday 08 December 2016 03:15, DFS wrote:

> dts= ['10-Mar-1998',
> '20-Aug-1997',
> '06-Sep-2009',
> '23-Jan-2010',
> '12-Feb-2010',
> '05-Nov-2010',
> '03-Sep-2009',
> '07-Nov-2014',
> '08-Mar-2013']
>
> Of course, the naive:
> min(dates) = '03-Sep-2009'
> max(dates) = '23-Jan-2010'
> is wrong.
>
> Not wanting to use any date parsing libraries, I came up with:

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

That's where you have gone wrong. By not using a date parsing library which
works and has been tested, you have to write your own dodgy and possibly buggy
parsing routine yourself.

Why reinvent the wheel?

> ========================================================================
> m=[('Dec','12'),('Nov','11'),('Oct','10'),('Sep','09'),
> ('Aug','08'),('Jul','07'),('Jun','06'),('May','05'),
> ('Apr','04'),('Mar','03'),('Feb','02'),('Jan','01')]
>
> #create new list with properly sortable date (YYYYMMDD)
> dts2 = []
> for d in dts:
> dts2.append((d[-4:]+dict(m)[d[3:6]]+d[:2],d))

Why do you convert m into a dict each and every time through the loop? If you
have a million items, you convert m to a dict a million times, and then throw
away the dict a million times. And then you'll complain that Python is slow...

m = {'Jan': 1, 'Feb': 2, 'Mar': 3, 'Apr': 4, 'May': 5, 'Jun': 6,
'Jul': 7, 'Aug': 8, 'Sep': 9, 'Oct': 10, 'Nov': 11, 'Dec': 12}

There you go, fixed that.

As for the rest, I haven't bothered to check your code. But here's a much
simpler, more Pythonic, way:

def date_to_seconds(string):
return time.mktime(time.strptime(string, '%d-%b-%Y'))


min(dts, key=date_to_seconds) # returns '20-Aug-1997'
max(dts, key=date_to_seconds) # returns '07-Nov-2014'


Works for at least Python 3.3 or better.

The best part of this is that you can now support any time format you like,
just by changing the format string '%d-%b-%Y'.



--
Steven
"Ever since I learned about confirmation bias, I've been seeing
it everywhere." - Jon Ronson

[Peter Heitzer]

I'd use strptime from the time module.

Then you could write
dts2.append(strptime(d,'%d-%b-%Y)
min and max return a struct_time type that can easily converted to
the original date format

--
Dipl.-Inform(FH) Peter Heitzer, peter....@rz.uni-regensburg.de

[DFS]

On 12/8/2016 12:16 AM, Steven D'Aprano wrote:
> On Thursday 08 December 2016 03:15, DFS wrote:
>
>> dts= ['10-Mar-1998',
>> '20-Aug-1997',
>> '06-Sep-2009',
>> '23-Jan-2010',
>> '12-Feb-2010',
>> '05-Nov-2010',
>> '03-Sep-2009',
>> '07-Nov-2014',
>> '08-Mar-2013']
>>
>> Of course, the naive:
>> min(dates) = '03-Sep-2009'
>> max(dates) = '23-Jan-2010'
>> is wrong.
>>
>> Not wanting to use any date parsing libraries, I came up with:
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> That's where you have gone wrong. By not using a date parsing library which
> works and has been tested, you have to write your own dodgy and possibly buggy
> parsing routine yourself.
>
> Why reinvent the wheel?

Because the "wheel" is a pain in the ass.

--------------------------------------------------------------
import time
dts=['10-Mar-1908','20-Aug-1937','06-Sep-1969','23-Jan-1952']

def date_to_seconds(string):
return time.mktime(time.strptime(string, '%d-%b-%Y'))

print min(dts, key=date_to_seconds)
--------------------------------------------------------------

OverflowError: mktime argument out of range

>> ========================================================================
>> m=[('Dec','12'),('Nov','11'),('Oct','10'),('Sep','09'),
>> ('Aug','08'),('Jul','07'),('Jun','06'),('May','05'),
>> ('Apr','04'),('Mar','03'),('Feb','02'),('Jan','01')]
>>
>> #create new list with properly sortable date (YYYYMMDD)
>> dts2 = []
>> for d in dts:
>> dts2.append((d[-4:]+dict(m)[d[3:6]]+d[:2],d))
>
>
> Why do you convert m into a dict each and every time through the loop? If you
> have a million items, you convert m to a dict a million times, and then throw
> away the dict a million times. And then you'll complain that Python is slow...

Python IS slow (for some things anyway).

But in this case, the list->dict conversion isn't; 1 million of them
takes just under 2 seconds.

---------------------------------------------------------------------------------
import sys,time,random

#lists
d= ['01','02','03','04','05','06','07','08','09','10']
d+=['11','12','13','14','15','16','17','18','19','20']
d+=['21','22','23','24','25','26','27','28']
m=['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec']
y,dt=[],[]
for yr in range(1900,2001):y.append(yr)


loops=int(sys.argv[1])

#populate list of random dates in format 'dd-MMM-yyyy'
start=time.clock()
for i in range(loops):
rd=random.choice(d)
rm=random.choice(m)
ry=random.choice(y)
dt.append(rd+'-'+rm+'-'+str(ry))
print "Build list of %s random dates: %.2g sec"
%(len(dt),time.clock()-start)
print


#create sortable 2nd list: convert list to dict
ml=
[('Dec','12'),('Nov','11'),('Oct','10'),('Sep','09'),('Aug','08'),('Jul','07')]
ml+=[('Jun','06'),('May','05'),('Apr','04'),('Mar','03'),('Feb','02'),('Jan','01')]
start=time.clock()
dts=[]
for d in dt:
dts.append((d[-4:]+dict(ml)[d[3:6]]+d[:2],d))
print "With list to dict conversion: %.2g sec" %(time.clock()-start)
print min(dts)[1], max(dts)[1]
print


#create sortable 2nd list: use dict
md={'Jan':'01','Feb':'02','Mar':'03','Apr':'04','May':'05','Jun':'06',
'Jul':'07','Aug':'08','Sep':'09','Oct':'10','Nov':'11','Dec':'12'}
start=time.clock()
dts=[]
for d in dt:
dts.append((d[-4:]+md[d[3:6]]+d[:2],d))
print "No list to dict conversion: %.2g sec" %(time.clock()-start)
print min(dts)[1], max(dts)[1]
---------------------------------------------------------------------------------

$ python temp.py 1000000
Build list of 1000000 random dates: 3.9 sec

With list to dict conversion: 2.7 sec
01-Jan-1900 28-Dec-2000

No list to dict conversion: 0.93 sec
01-Jan-1900 28-Dec-2000

> m = {'Jan': 1, 'Feb': 2, 'Mar': 3, 'Apr': 4, 'May': 5, 'Jun': 6,
> 'Jul': 7, 'Aug': 8, 'Sep': 9, 'Oct': 10, 'Nov': 11, 'Dec': 12}
>
> There you go, fixed that.

Broke it: "TypeError: cannot concatenate 'str' and 'int' objects"

This is what I need:
m = {'Jan':'01','Feb':'02','Mar':'03','Apr':'04','May':'05','Jun':'06',
'Jul':'07','Aug':'08','Sep':'09','Oct':'10','Nov':'11','Dec':'12'}

> As for the rest, I haven't bothered to check your code. But here's a much
> simpler, more Pythonic, way:
>
> def date_to_seconds(string):
> return time.mktime(time.strptime(string, '%d-%b-%Y'))
>
>
> min(dts, key=date_to_seconds) # returns '20-Aug-1997'
> max(dts, key=date_to_seconds) # returns '07-Nov-2014'
>
>
> Works for at least Python 3.3 or better.
>
> The best part of this is that you can now support any time format you like,
> just by changing the format string '%d-%b-%Y'.

I like that flexibility, but mktime is apparently useless for dates
prior to 'the epoch'.

[can...@gmail.com]

Am Donnerstag, 8. Dezember 2016 14:47:31 UTC+1 schrieb DFS:
> On 12/8/2016 12:16 AM, Steven D'Aprano wrote:
> > On Thursday 08 December 2016 03:15, DFS wrote:
> >
> >> dts= ['10-Mar-1998',
> >> '20-Aug-1997',
> >> '06-Sep-2009',
> >> '23-Jan-2010',
> >> '12-Feb-2010',
> >> '05-Nov-2010',
> >> '03-Sep-2009',
> >> '07-Nov-2014',
> >> '08-Mar-2013']
> >>
> >> Of course, the naive:
> >> min(dates) = '03-Sep-2009'
> >> max(dates) = '23-Jan-2010'
> >> is wrong.
> >>
> >> Not wanting to use any date parsing libraries, I came up with:
> > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> >
> > That's where you have gone wrong. By not using a date parsing library which
> > works and has been tested, you have to write your own dodgy and possibly buggy
> > parsing routine yourself.
> >
> > Why reinvent the wheel?
>
>
> Because the "wheel" is a pain in the ass.
>

Why? And why do you use this wording?

> --------------------------------------------------------------
> import time
> dts=['10-Mar-1908','20-Aug-1937','06-Sep-1969','23-Jan-1952']
> def date_to_seconds(string):
> return time.mktime(time.strptime(string, '%d-%b-%Y'))
> print min(dts, key=date_to_seconds)
> --------------------------------------------------------------
>
> OverflowError: mktime argument out of range
>
>

(snip)

>
> I like that flexibility, but mktime is apparently useless for dates
> prior to 'the epoch'.

With a little more experience in Python programming you should have discovered that time.mktime is not even required to do your calculations.

Please remove time.mktime from the function date_to_seconds and rename the function to date_to_timestruct.

-- Paolo

[Skip Montanaro]

Datetime has greater range I believe. I can't paste from my work computer,
but try this:

min(datetime.datetime.strptime(s, "%d-%b-%Y") for s in dts)

You should get the 1908 date instead of the 1969 date.

In general, you should always use datetime instead of time for doing date
manipulation and date arithmetic. It's a *whole lot better*.

Skip

[Cousin Stanley]

DFS wrote:

> ....

> Not wanting to use any date parsing libraries,

> ....

If you happen reconsider date parsing libraries
the strptime function from the datetime module
might be useful ....

#!/usr/bin/env python3

from datetime import datetime

dates = [ '10-Mar-1998' ,

'20-Aug-1997' ,
'06-Sep-2009' ,
'23-Jan-2010' ,
'12-Feb-2010' ,
'05-Nov-2010' ,
'03-Sep-2009' ,
'07-Nov-2014' ,
'08-Mar-2013' ]

dict_dates = { }

print( )

for this_date in dates :

dt = datetime.strptime( this_date , '%d-%b-%Y' )

sd = dt.strftime( '%Y-%m-%d' )

print( ' {0} .... {1}'.format( this_date , sd ) )

dict_dates[ sd ] = this_date


min_date = min( dict_dates.keys() )

max_date = max( dict_dates.keys() )

print( '\n {0} .... {1} min'.format( dict_dates[ min_date ] , min_date ) )

print( '\n {0} .... {1} max'.format( dict_dates[ max_date ] , max_date ) )








































--
Stanley C. Kitching
Human Being
Phoenix, Arizona