QDateTime::fromString() returns invalid Datetime
-
@kshegunov said in QDateTime::fromString() returns invalid Datetime:
Well, no. The date is still invalid - it doesn't (always) exist in "local time" no matter whether you force it after that.
You are correct that some strings representing a date/time are always invalid in some time zones, and not in others. My observation relates to the way QDateTime behaves, not the way time zones behave.
If you look at my example for Paris you will see the 2017 date is initially parsed, loaded and treated as Invalid (correctly). However, telling the QDateTime object containing the invalid time for Paris that it actually holds a UTC date/time generates the expected, valid result. The date/time components have been retained and the Invalid state is determined in light of the TimeSpec in use (LocalTime vs UTC). As the documentation for QDateTime::setTimeSpec() changing this results in a different point in time (in this case a valid one).
Note, that this is distinctly not the same as converting the same point in time time between time zones. Attempting to determine the corresponding UTC time for an Invalid Paris time, in this case by adjusting by an undefined 1 or 2 hours, should also be expected to be Invalid or exhibit undefined behaviour. Interestingly, Qt loaded the original date with an offsetFromUtc() == 0 (there being no valid offset corresponding to the date string in Paris) and behaves as such when asked for toUTC().
I work in aviation where everything is UTC/Zulu time by default, and local time by exception, and I agree with you that unambiguous date/time representations are important. Appending "Z" or "+00:00" to incoming unqualified date strings is an option, but having a source that issues unambiguous data from the start is definitely preferable.
-
@ChrisW67 said in QDateTime::fromString() returns invalid Datetime:
I work in aviation where everything is UTC/Zulu time by default
Hallelujah! I too have worked where multiple timezones are involved, and it's a pain. Why, oh why, can't the whole world agree to just use UTC everywhere, no time zones, it would make programs (and appointments) so much easier....?
-
@JonB said in QDateTime::fromString() returns invalid Datetime:
Hallelujah! I too have worked where multiple timezones are involved, and it's a pain.
It is not all roses though :( Times for passenger schedules are local, flight plans are all Zulu. In aviation runway lengths and widths are in metres, altitudes in feet, sometimes metres, referenced to the actual barometric pressure at the aerodrome or in the area (in hectopascals or inches of Mercury), or a flight level (hundreds of feet) relative to a standard pressure depending on the region and height at which you are flying. Airspeed is in knots or a Mach number, distances in nautical miles. Bearings are in degrees, typically magnetic, but North is sometimes 0 sometimes 360. Aircraft weight is given in pounds or kilogrammes/tonnes, engine thrust in pound force or newtons, fuel in pound, litres, kg/tonnes and sometime even gallons (US or imperial). Terrain elevation might be given in feet or metres, but safety clearances (above or laterally) are in metres for terminal procedure design and feet in general elsewhere. Temperatures are in Celsius mostly, Fahrenheit sometimes. Weather reports at aerodromes give horizontal visibility in kilometres or statute miles. Latitudes/longitudes are sometimes presented in decimal, sometimes sexagesimal with variable precision/spacing/formatting. Aircraft structures are variously metric or imperial.
-
@ChrisW67
OMG!!But my desire is that there is no "local time" anywhere, just everyone uses UTC always. Miles simpler....
The crazy things about all those different units you mention is that --- once they have to proper code to measure/convert --- computers really don't have any problems dealing with them all effortlessly. Which is why pilots' days --- like many other professions --- are numbered... :)
-
@ChrisW67 said in QDateTime::fromString() returns invalid Datetime:
You are correct that some strings representing a date/time are always invalid in some time zones, and not in others. My observation relates to the way QDateTime behaves, not the way time zones behave.
It's worse than that. Some points are ambiguous in local time with DST in effect, and some plainly don't exist at all. Not to mention you're suggesting a hidden use case for
QDateTime
, which may or may not work due to an implementation detail. Sorry, it simply is an incorrect solution. -
This post is deleted!
-
@Christian-M
Hello and welcome.I do not understand. The whole point is that is that from 02:00 to 02:59 on that day there IS no "valid time". It does not exist. In local time, which is what you are asking about. At best it is ambiguous: does 02:30 represent 30 minutes after the clocks changed from 02:00 or 30 minutes before they changed to 03:00? So I don't know what you mean by "To get a valid QDateTime for example in UTC" when the local time is simply not valid. If you want to treat 02:30 as a UTC time that is fine, but is quite a different time from local time.
-
This post is deleted!
-
@Christian-M
OK, so the only correct way to do this is: convert from foreign TZ (where 02:00--03:00 is valid) to UTC, convert from UTC to your local TZ (no ambiguity, a given UTC only maps to one local TZ time). It never was correct to even try to interpret that foreign TZ time as a local time. You might as well interpret that foreign time as though it were UTC, since you don't know anyway, then at least it won't cause an error.QDateTime QDateTime::fromString(const QString &string, const QString &format, const QTimeZone &timeZone);
Indeed! Can one not achieve this by either adding some TZ information to the string or is void QDateTime::setTimeZone(const QTimeZone &toZone) just what this is for? I might have a play later if I have some time....
P.S.
I see that @Christian-Ehrlicher said earlierBut where do you pass this information? QDateTime::fromString() is using the current locale. Use QLocale::toDateTime() instead
Maybe that will allow conversion from a foreign local time?
-
This post is deleted!
-
@Christian-M
Firstly, your post has literally crossed with my adding a P.S. to my reply above.foreingTz.setTimeZone(QTimeZone(QByteArray("UTC+3")));
Are you sure this approach is correct (I don't think so)? Ignoring for the moment that you may know the foreign timezone name but not its UTC offset. Let's say this foreign TZ has its own daylight savings rules, and let's say that this foreign local time does (or does not, you don't know) lie in its DST. Then, correct me if I am wrong, I cannot see how from
UTC+3
datetime code could possibly know whether DST is or is not in effect on the stated time. In effect, you need to know whether to pass, say,UTC+3
vsUTC+4
for that foreign local time at its different times of year, and you do not know that. In which case, it cannot convert correctly.....???[P.S. I assume your input string does not itself come with any
UTC+...
suffix on it. Obviously if it does then the whole thing is trivial, since one can get to UTC from that and then UTC to your local. I assume you mean the time string has no UTC information on it, just that you know it comes from e.g. Europe in local time and are wanting to convert to Melbourne local time.] -
This post is deleted!
-
@Christian-M
Firstly, I mistook you for the OP, I see now you were a responder. I was trying to answer/help the OP. I am not the one who is confused! I have been working with cross-timezone code for decades! What I am trying to discover is how to do something in Qt with this, which I have not done.My understanding of what the OP was asking is:
- He is in Australia/Melbourne, that is his local tz.
- He gets a datetime string to parse. It does not have a tz specification, e.g. something like
"2017-03-26T02:14:34.000"
. - He does know it represents a local time, and he does know which timezone it is from, but it is not his local timezone. And he does not know whether the foreign time lies in foreign DST or Standard time, he just knows the time is "correct" for a foreign local time. This is important.
- The question is what Qt code can he write to correctly parse that datetime from its "foreign" tz to a datetime in his local tz (or to UTC).
I do not see how your
QCalendar cal(QCalendar::System::Julian); QDateTime ts = QDateTime::fromString("02/10/2022 02:00:00", "dd/MM/yyyy hh:mm:ss", cal); //valid
can possibly address this. The fact that you use Julian calendar does not alter the problem. Since nowhere have you told it that the string comes from a foreign local time, it has no idea what this "02:00" time actually was in UTC.
Could you please show a worked example you claim which addresses this correctly? Let's say you are in Australia and you are given
"02/10/2022 02:00:00"
which you know to be in, say, UK time. But your code is not running in UK. Please show how you convert that to, say, UTC time in Qt. Thank you. -
This post is deleted!
-
@Christian-M
Hi, it has taken me a day to get back to you.Unfortunately I am on Qt 5.12 where there is no
QCalendar
, that was introduced at 5.14. So I cannot test your whole principle of a default Julian calendar and passing that toQDateTime::fromString()
. So unless you know of an equivalent to your code for Qt <= 5.12 I cannot test anything.What I can say is that at 5.12
QDateTime foreingTz = QDateTime::fromString(datetime_foreign_with_no_timezone_spec); foreingTz.setTimeZone(QTimeZone(QByteArray(foreign_timezone)));
does seem to work, where my objective is to treat
datetime_foreign_with_no_timezone_spec
as a local time inforeign_timezone
. What it does is:- Initial
fromString()
treatsdatetime_foreign_with_no_timezone_spec
as a local time in my local timezone. - Calling
setTimeZone(foreign_timezone)
on that does seem to alter the interpretation of the original datetime string as a local time in the foreign timezone (which is what I am trying to achieve).
- Initial