Question

I am trying to get all files for last 12 hours,
file names has following format %Y-%m-%d %H

Here is my python script, i am trying to make work

last12HourDateTime = datetime.today() - timedelta(hours = 12)
allowedFormat = last12HourDateTime.strftime('%Y-%m-%d %H')


for filePath in glob.glob(allowedFormat):

I know there are dozens ways doing it, yet i would like to know if its possible this way

(EDIT)
I was able to complete it by

allowedFormats =[]
for i in range (1,12):
    last12HourDateTime = datetime.today() - timedelta(hours = (i - 1))
    allowedFormats.append(last12HourDateTime.strftime('%Y-%m-%d-%H.log'))


for allowedFormat in allowedFormats:
    for filePath in glob.glob(allowedFormat):

Yet still looking for better efficient solution

Answer 1

Technically: yes.

Worth it: No.

Reason: Regex has no understanding of numeric value and thus can’t do arithmetical comparisons (x > z – 12).

In other words: You’d have to generate a custom regex for every single use and would thus be way better off doing it with a real date format parser and a date class that’s capable of what you’re after, as in regex you’d have to produce huge amounts of ANDed (...|...) groups and pretty much end up with basic batched string comparisons (which would rechnically still be valid regexes, yet lack any of its higher purposes).

---

Most text related questions of the pattern “Is x possible in regex?” can technically be answered with YES. (see above)

Thus I’d much rather ask: “Should I (try to) do x with regex?” or “Is regex the right tool for x?”.

If your only tool is a hammer…

---

If you wanted to at least narrow down the list of potential matches (before doing any real date arithmetic) you’d have to generate a regex based on these rules (from the top of my head, no guarantees)

(I will use h for current hour, d for current day, m for current month and y for current year.)

if (h < 12)
    %dh = '(?:yesterday (?:1[2-9]|2[0-3])|today [0-9]{1,2})'
else
    %dh = '(?:tomorrow (?:[0-9]|1[0-1])|today [0-9]{1,2})'

if (d == 1)
    %m = '(?:lastmonth|thismonth)'
else if (d == 31 && count of days in m == 31 ||
         d == 30 && count of days in m == 30 ||
         m == 2 && d == 28 ||
         m == 2 && d == 29 && y is leap year)
    %m = '(?:thismonth|nextmonth)'
else
    %m = 'thismonth'

if (m == 1)
    %y = '(lastyear|thisyear)'
else if (m == 12)
    %y = '(?:thisyear|nextyear)'
else
    %y = 'thisyear'

where you’d replace yesterday, thisyear, etc with their respective numeric values.

And form a regex of pattern %y-%m-%dh where you’d replace %y, %m , %dh with the values as determined above.

Again: date arithmetic is tricky, thus my algo above probably contains mistakes.

---

I don’t know the wider context of your question so I can only guess.
Based on the information you gave (and assuming that the file names/files don’t change by 100% for every search, thus allowing some degree of caching), I’d probably go with something like this:

Enumerate your list of files and convert their date-formatted filenames into UNIX timestamps, adding each of them them to a list (possibly better: create container objects holding timestamp & filepath, otherwise you’d have to geceive the filepath by converting the timestamp back into a date formatted string and require the hierarchy to be flat). Sort the list. Get range of matching files using a modified binary search (in which instead of searching for an actual value match you search for a range of relative matches. I have no sample code right now, but it’s not that difficult).

Now assuming that from time to time files get added/deleted you’d either have to be able to monitor those system events and update your list.

Creating the list the first time takes O(n) (+ O(nlogn) for the sorting) but if you’re able to smartly update your cached timestamp list you should be able to gain quite some performance.