[GNC] Working with dates in Postgresql DB
John Ralls
jralls at ceridwen.us
Fri May 1 14:00:37 EDT 2020
David,
You're not thinking it through: It's about 11:00 on Friday 1 May here in California but it's 03:00 on Saturday 2 May in Western Australia. Chopping off the time doesn't solve anything, a point illustrated by Finfort when he pointed out that just changing the time on the errant dates would put them in the wrong day.
Regards,
John Ralls
> On Apr 30, 2020, at 10:24 PM, D. <sunfish62 at yahoo.com> wrote:
>
> Thanks for the reply. I do understand the challenges this poses-- both from the perspective of managing it on a daily basis, and from that of the difficulty of changing the underlying system. At least, conceptually!
>
> Is there any option to simply ignore the time portion in these timestamps? It would seem to me that one could focus on that, and simplify the process piece by piece. Of course, not being a programmer, I'm just a silly voice in the wilderness.
>
> David
>
>
> -------- Original Message --------
> From: John Ralls <jralls at ceridwen.us>
> Sent: Fri May 01 10:32:09 GMT+05:30 2020
> To: "D." <sunfish62 at yahoo.com>
> Cc: "finfort at gmail.com" <finfort at gmail.com>, "D. via gnucash-user" <gnucash-user at gnucash.org>
> Subject: Re: [GNC] Working with dates in Postgresql DB
>
> David,
>
> I don't know why the decision was made to use time, it was taken long before I joined the project, but it probably has to do with that being the way computers keep time, in UTC and the accompanying date-time manipulation functions in the C standard library were readily available. Over the years various developers have piled more manipulation functions on top of it, or added other libraries to the side because they made doing something easier, and splattered it all over the code so that changing the base design would involve chasing down and reworking all of those disparate functions. As I said, no one has expressed much enthusiasm for taking it on.
>
> Knowing now that using time instead of date was a poor decision is just applying 20/20 hindsight to criticize Linas's decision 22 years ago. It can't change it. I can't say that I would have decided differently.
>
> Regards,
> John Ralls
>
>
>> On Apr 30, 2020, at 8:46 PM, D. <sunfish62 at yahoo.com> wrote:
>>
>> John,
>>
>> I somewhat remember the discussion back in 2011 about the timestamp, but do not recall all the details. Remind me why it is that GnuCash uses timestamps in these date fields? All these steps and workarounds that take place to present the proper date around the world.
>>
>> Wouldn't it be simpler just to store a bare date?
>>
>> Or just ignore the time portion and drop the conversion between UTC and locale time altogether?
>>
>> Everyone refers to them as dates ("transaction date" "invoice date" "posting date"). The software arbitrarily uses the same time for everything in a futile attempt to properly display dates for all timezones (the solution in this thread underscores that fact, insofar as you are recommending the user to arbitrarily change all other times to the "standard").
>>
>> Of what use is it to store the added detail of an arbitrary timestamp in a field that is treated everywhere as a date? What is gained?
>>
>> David T.
>>
>>
>> -------- Original Message --------
>> From: John Ralls <jralls at ceridwen.us>
>> Sent: Fri May 01 00:48:57 GMT+05:30 2020
>> To: "finfort at gmail.com" <finfort at gmail.com>
>> Cc: Gnucash Users <gnucash-user at gnucash.org>
>> Subject: Re: [GNC] Working with dates in Postgresql DB
>>
>> GnuCash stores all dates as UTC but displays them as local, applying the timezone rules for the date, not for today. So in EEST 2020-02-12 22:00:00 displays as 2020-02-13, 2020-06-12 21:00:00 displays as 2020-06-13, but 2020-02-21 21:00:00 displays as 2020-02-21.
>>
>> Regards,
>> John Ralls
>>
>>
>>> On Apr 30, 2020, at 11:42 AM, finfort at gmail.com wrote:
>>>
>>> It is not just adding one day, it depends on the time.
>>>
>>> Looks like time 00:00:00 is the same date, not next.
>>>
>>> From 21:00:00 is the next date in most cases, but I did not check all transactions manually =)
>>>
>>> How the program converts this wrong dates to the correct ones in its GUI?
>>>
>>> I believe I have found a correct way to convert all the dates including wrong ones to correct dates in Postgresql (pgAdmin 4):
>>>
>>>
>>> date(t.post_date AT TIME ZONE 'UTC' AT TIME ZONE 'EEST') AS DATE_AT_timezone_EEST
>>>
>>> EEST is a correct zone in my case. CEST does not work.
>>>
>>> The transactions.post_date type is timestamp without timezone: 2017-12-31 21:00:00
>>>
>>> t.post_date AT TIME ZONE 'EEST' AS timestamp_AT_timezone_EEST gives 2017-12-31 21:00:00+3
>>>
>>> date(t.post_date AT TIME ZONE 'UTC' AT TIME ZONE 'EEST') AS DATE_AT_timezone_EEST gives 2018-01-01
>>>
>>> Looks strange but works.
>>>
>>> 2.
>>>
>>> There are only 3 transactions with 22:00:00 not connected with invoices.
>>>
>>> There is only 1 transaction with 21:00:00 not connected with invoices.
>>>
>>> Thinking how to find them...
>>>
>>>
>>>
>>> On 30/04/2020 21:25, John Ralls wrote:
>>>> Hmm, true. Should be always, since you're in a time zone east of the prime meridian. So you also want to increment the day on those. I think it would be safest to do it in two queries, the first one being
>>>>
>>>> update transactions post_date = post_date + interval '1 day' where post_date::TIME != '10:59:00';
>>>>
>>>> and the second to update the time as before.
>>>>
>>>> Regards,
>>>> John Ralls
>>>>
>>>>
>>>>> On Apr 30, 2020, at 11:07 AM, Finfort <finfort at gmail.com> wrote:
>>>>>
>>>>> If post_date is 2017-12-31 22:00 or 23:00, it means (sometimes?) the real date is 2018-01-01. At least in cases where I manually checked the invoices.
>>>>> Setting all times to 10:59:00 will give wrong dates in the program.
>>>>> Now they are displayed correctly in the program somehow...
>>>>>
>>>>>
>>>>>
>>>>>> On Apr 30, 2020 at 20:59, <John Ralls> wrote:
>>>>>>
>>>>>> I don't think that's necessary.
>>>>>>
>>>>>> To fix the wrong times just do an update query, something like
>>>>>>
>>>>>> update transactions set post_date::TIME = 10:59:00 where post_date::TIME != 10:59:00;
>>>>>>
>>>>>> I don't know Postgresql's date-time functions well enough to know if that syntax works, you might have to adjust it a bit. You might create a table with a DATETIME column and put a couple of rows in it to test against while you tweak. Make sure that GnuCash isn't connected to the database when you run that.
>>>>>>
>>>>>> Regards,
>>>>>> John Ralls
>>>>>>
>>>>>>
>>>>>>> On Apr 30, 2020, at 10:45 AM, Finfort <finfort at gmail.com> wrote:
>>>>>>> How can I help?
>>>>>>> I can send you my gnucash file if it helps to find all the bugs.
>>>>>>> And how can I change now my wrong dates in transactions?
>>>>>>>
>>>>>>>> On Apr 30, 2020 at 20:41, <John Ralls> wrote:
>>>>>>>> Yeah, it's definitely a bug. I easily found the wrong code and I'll fix it for 3.903 and 3.11.
>>>>>>>> The query actually accounts for only 543 of the 547 wrong times, so there's another error somewhere else.
>>>>>>>> Regards,
>>>>>>>> John Ralls
>>>>>>>>
>>>>>>>>> On Apr 30, 2020, at 10:27 AM, Finfort <finfort at gmail.com> wrote:
>>>>>>>>> Also I tried to unpost and post again. No changes.
>>>>>>>>>
>>>>>>>>>> On Apr 30, 2020 at 19:44, <Finfort> wrote:
>>>>>>>>>> Hi John,
>>>>>>>>>> The result is:
>>>>>>>>>> 22:00:00 253
>>>>>>>>>> 00:00:00 18
>>>>>>>>>> 21:00:00 250
>>>>>>>>>> 23:00:00 22
>>>>>>>>>> So wrong dates are only when I use invoices.
>>>>>>>>>> On 29/04/2020 23:56, John Ralls wrote:
>>>>>>>>>>> Please remember to copy the list on all replies.
>>>>>>>>>>>
>>>>>>>>>>> I take it that that means that you do in fact use the business invoice features. Let's see if that's the source of the problem. Run this query:
>>>>>>>>>>>
>>>>>>>>>>> select t.post_date::TIME count(t.post_date::TIME) from transactions as t inner join invoices as i on i.post_txn = t.guid group by t.post_date::TIME;
>>>>>>>>>>>
>>>>>>>>>>> Regards,
>>>>>>>>>>> John Ralls
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>> On Apr 29, 2020, at 1:41 PM, Finfort <finfort at gmail.com> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>> But the program use business features like entering invoices or bills.
>>>>>>>>>>>> And we have this mess.
>>>>>>>>>>>> How we can manage that?
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>> On Apr 29, 2020 at 23:23, <John Ralls> wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>> Dimon,
>>>>>>>>>>>>>
>>>>>>>>>>>>> I'm in Silicon Valley, so 10 hours behind you.
>>>>>>>>>>>>>
>>>>>>>>>>>>> By "simple invoices" do you mean that some of the transactions are created using business invoices?
>>>>>>>>>>>>>
>>>>>>>>>>>>> My book goes back to 1993 and entry_dates begin in 2001. We changed the transaction time from local midnight in early 2011 so `select distinct time(post_date) from transactions;` returns
>>>>>>>>>>>>> 10:59:00
>>>>>>>>>>>>> 07:00:00
>>>>>>>>>>>>>
>>>>>>>>>>>>> If I say instead `select distinct time(post_date) from transactions where post_date > '2011-01-01';` I just get 10:59:00.
>>>>>>>>>>>>>
>>>>>>>>>>>>> But I don't use the business features, so if that's the problem my book won't show it.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Regards,
>>>>>>>>>>>>> John Ralls
>>>>>>>>>>>>>
>>>>>>>>>>>>>> On Apr 29, 2020, at 10:13 AM, Finfort <finfort at gmail.com> wrote:
>>>>>>>>>>>>>> Hi John!
>>>>>>>>>>>>>> You are here finally!
>>>>>>>>>>>>>> Waiting for you all the day :)
>>>>>>>>>>>>>> All my data I have entered inside Gnucash 3.7, Ubuntu. No imports! Scheduled are ok!
>>>>>>>>>>>>>> Just simple invoices inside the program!
>>>>>>>>>>>>>> The SQL type conversions inside Postgres give better results with 22:00 but 21:00 show the same date again even in April - summer time where is for example 2018-06-04 21:00:00+03.
>>>>>>>>>>>>>> 22:00+02 is 00:00 of the next day, 21:00+03 (summer time) is 00:00 of the next day but conversion does not work...
>>>>>>>>>>>>>> So, maybe you could try this SQL to check your records and revise the procedure which posts the data to DB?
>>>>>>>>>>>>>> Thank you,
>>>>>>>>>>>>>> Dimon.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On Apr 29, 2020 at 19:50, <John Ralls> wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On Apr 29, 2020, at 2:18 AM, finfort at gmail.com wrote:
>>>>>>>>>>>>>>>> Dear John,
>>>>>>>>>>>>>>>> Thank you for your response.
>>>>>>>>>>>>>>>> I have collected some statistics from my DB.
>>>>>>>>>>>>>>>> My DB has 1724 records - transactions.
>>>>>>>>>>>>>>>> This is my SQL query, it is pretty simple and shows all the combinations of times in posted_date timestamps in transactions table, number of repetitions for that time value, min enter_date, max enter_date:
>>>>>>>>>>>>>>>> SELECT
>>>>>>>>>>>>>>>> t.post_date::TIME as "POST TIME",
>>>>>>>>>>>>>>>> COUNT(t.post_date::TIME) as "REPS",
>>>>>>>>>>>>>>>> min(t.enter_date) as "MIN ENTER DATE",
>>>>>>>>>>>>>>>> max(t.enter_date) as "MAX ENTER DATE"
>>>>>>>>>>>>>>>> FROM transactions t
>>>>>>>>>>>>>>>> GROUP BY t.post_date::TIME
>>>>>>>>>>>>>>>> ORDER BY t.post_date::TIME
>>>>>>>>>>>>>>>> Here are the results:
>>>>>>>>>>>>>>>> ----
>>>>>>>>>>>>>>>> POST TIME REPS MIN ENTER DATE MAX ENTER DATE
>>>>>>>>>>>>>>>> "00:00:00" 18 "2020-01-26 18:07:14" "2020-01-28 19:11:07"
>>>>>>>>>>>>>>>> "10:59:00" 1177 "2019-12-23 17:55:29" "2020-04-23 11:24:24"
>>>>>>>>>>>>>>>> "21:00:00" 251 "2020-01-08 17:43:54" "2020-04-23 10:36:33"
>>>>>>>>>>>>>>>> "22:00:00" 256 "2020-01-08 17:06:59" "2020-04-23 11:24:08"
>>>>>>>>>>>>>>>> "23:00:00" 22 "2020-01-27 19:16:04" "2020-01-28 19:39:49"
>>>>>>>>>>>>>>>> ----
>>>>>>>>>>>>>>>> I live in Cyprus, here is UTC +2 and summer time UTC +3, as I know.
>>>>>>>>>>>>>>>> I started to study Gnucash in December 2019 and have entered my data of 2016-2020.
>>>>>>>>>>>>>>>> I never changed my place and time zone in the period of working with Gnucash.
>>>>>>>>>>>>>>>> 1. Most of the records have time in date_posted 10:59:00 for all the period of data entering.
>>>>>>>>>>>>>>>> 2. Only 2 days of entering have the results of 00:00:00 - 18 records.
>>>>>>>>>>>>>>>> 3. Only 2 days of entering have the results of 23:00:00 - 22 records.
>>>>>>>>>>>>>>>> 4. 21:00:00 and 22:00:00 - 500+ records - 30% of transactions for all the period of data entering.
>>>>>>>>>>>>>>>> Can you please explain that?
>>>>>>>>>>>>>>>> Why I have so many different time stamps? When and why the system decides to write time different from 10:59:00?
>>>>>>>>>>>>>>>> I understand that the system writes real ENTERING date and time and it is reasonable to use the time zone somehow.
>>>>>>>>>>>>>>>> When I POST the document with exact date in it I suppose to see this POST DATE the same wherever in Cyprus or UK or USA. But entering the same date I can have 5 different results. How it works and what is the reason - I have no idea...
>>>>>>>>>>>>>>>> Maybe you can give some examples and the algorithm to convert these dates? Where else I have to convert dates?
>>>>>>>>>>>>>>> As those are all posted dates you've found a bug or two as posted date should always have a 10:59:00 timestamp. The 21:00 and 22:00 times are clearly midnight local, and which one is used *should* be determined by whether DST is in effect for the posted date in your locale. It seems that 40 transactions somehow used UK time instead of Cypress time.
>>>>>>>>>>>>>>> Did you enter all of the transactions from the GnuCash UI or did you import some of them? If you imported some is there any way to tell which were imported (and from where and by what method), perhaps by the accounts their splits are in or because you still have some of the import files?
>>>>>>>>>>>>>>> Were any of them created by scheduled transactions?
>>>>>>>>>>>>>>> Regards,
>>>>>>>>>>>>>>> John Ralls
>>>>>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>
>>>>>>
>>
>> _______________________________________________
>> gnucash-user mailing list
>> gnucash-user at gnucash.org
>> To update your subscription preferences or to unsubscribe:
>> https://lists.gnucash.org/mailman/listinfo/gnucash-user
>> If you are using Nabble or Gmane, please see https://wiki.gnucash.org/wiki/Mailing_Lists for more information.
>> -----
>> Please remember to CC this list on all your replies.
>> You can do this by using Reply-To-List or Reply-All.
>>
>
>
More information about the gnucash-user
mailing list