KLocalizedString regression
Hi all, ki18ncp(Units, %1 second, %1 seconds).subs(1.1, 0, 'f', 1).toString() used to work (return 1.1 seconds) but now it returns (1.1 seconds(I18N_PLURAL_ARGUMENT_MISSING)) which screws up my graphing widget. This causes my unit tests to fail. Can someone fix this pretty please? I can't work around this. John
Re: KLocalizedString regression
On 25 March 2011 12:45, Chusslove Illich caslav.i...@gmx.net wrote: [: John Tapsel :] ki18ncp(Units, %1 second, %1 seconds).subs(1.1, 0, 'f', 1).toString() used to work (return 1.1 seconds) but now it returns (1.1 seconds(I18N_PLURAL_ARGUMENT_MISSING)) which screws up my graphing widget. This is the way it (should have) always worked. The plural call expects to get an integer, and if does not, it shows this warning (in debug build, not in release build). That's not good :-( If the number is small, I want to show 2 seconds but if it's large I want like 2.2e4 seconds. You're saying that my API needs to require the user to now pass *two* strings for that? On the language level, plural declination kicks in only on integer values, while real values always get plural form. This seems to hold for all languages currently known in KDE. In English, all of 1.0 seconds, 1.1 seconds, 1.3 seconds, and 2.3 seconds, are (?) both correct and expected forms, as opposed to 1 second and 2 seconds. And that's fine. So do that. But don't stick a big warning there! If I specify a precision of 0, then it should be 1 second. If the precision is 1 then 1.0 seconds is fine. John
Re: KLocalizedString regression
[: John Tapsell :] If the number is small, I want to show 2 seconds but if it's large I want like 2.2e4 seconds. You're saying that my API needs to require the user to now pass *two* strings for that? Well, yes. Two grammar-wise different text fragments (enumerable vs. measurable quantities) are being used in alternation, so each should be represented by its own string. The author of KUnitConversion wanted the same feature, so there too we ended up defining two strings, e.g: setDefaultUnit(UP(Meter, 1, ... ki18nc(amount in units (real), %1 meters), ki18ncp(amount in units (integer), %1 meter, %1 meters) )); (from kdelibs/kunitconversion/length.cpp). In my opinion, however, this alternation is bad style-wise. The display should use either the measurable form throughout (2.0 seconds; 2.2e4 seconds), or the enumerable form throughout (2 seconds; 6 hours 6 minutes). -- Chusslove Illich (Часлав Илић) signature.asc Description: This is a digitally signed message part.
Re: KLocalizedString regression
On 25 March 2011 13:19, Chusslove Illich caslav.i...@gmx.net wrote: [: John Tapsell :] If the number is small, I want to show 2 seconds but if it's large I want like 2.2e4 seconds. You're saying that my API needs to require the user to now pass *two* strings for that? Well, yes. Two grammar-wise different text fragments (enumerable vs. measurable quantities) are being used in alternation, so each should be represented by its own string. The author of KUnitConversion wanted the same feature, so there too we ended up defining two strings, e.g: setDefaultUnit(UP(Meter, 1, ... ki18nc(amount in units (real), %1 meters), ki18ncp(amount in units (integer), %1 meter, %1 meters) )); Except that in my case, I'm forcing this on the users of my API. Just to get around this! Can't you see how crazy this is? In my opinion, however, this alternation is bad style-wise. The display should use either the measurable form throughout (2.0 seconds; 2.2e4 seconds), or the enumerable form throughout (2 seconds; 6 hours 6 minutes). But what if I want to allow both? This is for labelling the axis of a graph. I now have to have an API and forces the users to give the same string twice.
Re: KLocalizedString regression
On 25 March 2011 13:59, John Tapsell johnf...@gmail.com wrote: On 25 March 2011 13:19, Chusslove Illich caslav.i...@gmx.net wrote: [: John Tapsell :] If the number is small, I want to show 2 seconds but if it's large I want like 2.2e4 seconds. You're saying that my API needs to require the user to now pass *two* strings for that? Well, yes. Two grammar-wise different text fragments (enumerable vs. measurable quantities) are being used in alternation, so each should be represented by its own string. The author of KUnitConversion wanted the same feature, so there too we ended up defining two strings, e.g: setDefaultUnit(UP(Meter, 1, ... ki18nc(amount in units (real), %1 meters), ki18ncp(amount in units (integer), %1 meter, %1 meters) )); Except that in my case, I'm forcing this on the users of my API. Just to get around this! Can't you see how crazy this is? In my opinion, however, this alternation is bad style-wise. The display should use either the measurable form throughout (2.0 seconds; 2.2e4 seconds), or the enumerable form throughout (2 seconds; 6 hours 6 minutes). But what if I want to allow both? This is for labelling the axis of a graph. I now have to have an API and forces the users to give the same string twice. Chusslove, I just had an idea - so okay, if I'm forced to work around this feature, could I do: string.remove((I18N_PLURAL_ARGUMENT_MISSING)); This string won't change, and isn't translated right? John
Re: KLocalizedString regression
On 25 March 2011 14:20, Thiago Macieira thi...@kde.org wrote: Em sexta-feira, 25 de março de 2011, às 13:59:11, John Tapsell escreveu: Well, yes. Two grammar-wise different text fragments (enumerable vs. measurable quantities) are being used in alternation, so each should be represented by its own string. The author of KUnitConversion wanted the same feature, so there too we ended up defining two strings, e.g: setDefaultUnit(UP(Meter, 1, ... ki18nc(amount in units (real), %1 meters), ki18ncp(amount in units (integer), %1 meter, %1 meters) )); Except that in my case, I'm forcing this on the users of my API. Just to get around this! Can't you see how crazy this is? No. You haven't said anything about your API yet. Maybe the problem is in the API. But we don't know, because you haven't included a link to what API you're talking about. I have a widget, KSignalPlotter, which plots data points. The usage is: KSignalPlotter plotter; plotter.setUnit( ki18ncp(Units, %1 second, %1 seconds) ); plotter.addBeam(Qt::red); plotter.addSample( QListqreal() 2 ); For example. The axis will then be labelled using the unit string. The precision will be calculated automatically. Huh? Why do you want to show 1.0 seconds in the axis? Shouldn't you have 1.0 in the axis and then seconds in the axis label? Because typically we have a lot more horizontal space than vertical space. Adding room for axis labels would take up valuable vertical space. John
Re: KLocalizedString regression
Em sexta-feira, 25 de março de 2011, às 14:39:22, John Tapsell escreveu: I have a widget, KSignalPlotter, which plots data points. The usage is: KSignalPlotter plotter; plotter.setUnit( ki18ncp(Units, %1 second, %1 seconds) ); plotter.addBeam(Qt::red); plotter.addSample( QListqreal() 2 ); For example. The axis will then be labelled using the unit string. The precision will be calculated automatically. So you might have a graph showing 1, 2, and 3 where the unit is 2.2e4 seconds ? Shouldn't you show instead 2.2e4, 4.4e4 and 6.6e4 with a unit of seconds ? -- Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org Senior Product Manager - Nokia, Qt Development Frameworks PGP/GPG: 0x6EF45358; fingerprint: E067 918B B660 DBD1 105C 966C 33F5 F005 6EF4 5358 signature.asc Description: This is a digitally signed message part.
Re: KLocalizedString regression
Em sexta-feira, 25 de março de 2011, às 15:17:33, John Tapsell escreveu: The graph would show 1 second 2 seconds 3 seconds on the axis in this case. IMHO it should show: 1 s 2 s 3 s -- Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org Senior Product Manager - Nokia, Qt Development Frameworks PGP/GPG: 0x6EF45358; fingerprint: E067 918B B660 DBD1 105C 966C 33F5 F005 6EF4 5358 signature.asc Description: This is a digitally signed message part.
Re: KLocalizedString regression
[: John Tapsell :] Except that in my case, I'm forcing this on the users of my API. Just to get around this! Can't you see how crazy this is? An integer is mandatory to decide the plural form, a float cannot be used, due to the following. Suppose there are m plural forms for a given language. The plural call pushes the integer through a formula, defined by that language (by each PO file to be precise), which maps it to 0 through m - 1. If the plural call would get a float instead, how should it decide which of the m forms to map to? For example, in my language forms are 0 through 3, and the correct one for a floating value is 2 -- not 0, 1, or 3. On the implementation level, what happens now when you do not supply an integer, is that non-plural call is simply passed down to non-plural gettext() from libc. It is a fluke that for English it silently returns the plural, and not the singular string. For all I know, if the libc were built in develop/debug mode (I use distribution package), it could do something else. Therefore two independent KLocalizedString parameters are indeed necessary for the API, if you want to keep the integer/real alternation. Since it is about graph plotting, I suggest that you make the API such that the real- value string is mandatory, and the integer-value string optional. I.e. let the client choose if he wants alternation or not. Furthermore, I would suggest that the axis tick labels be pure numbers, and the axis name and unit shown separate in one place, e.g: -- 0 1 2 3 4 5 6 time [seconds] This is how graph axes are normaly presented, at least in engineering fields, and it would sidestep the complete issue. (I realize that this suggestion is off-topic as such for this thread, but I couldn't resist since I have strong feelings about proper graph plotting.) -- Chusslove Illich (Часлав Илић) signature.asc Description: This is a digitally signed message part.
Re: KLocalizedString regression
On 25 March 2011 21:42, Chusslove Illich caslav.i...@gmx.net wrote: [: John Tapsell :] Except that in my case, I'm forcing this on the users of my API. Just to get around this! Can't you see how crazy this is? An integer is mandatory to decide the plural form, a float cannot be used, due to the following. Suppose there are m plural forms for a given language. The plural call pushes the integer through a formula, defined by that language (by each PO file to be precise), which maps it to 0 through m - 1. If the plural call would get a float instead, how should it decide which of the m forms to map to? For example, in my language forms are 0 through 3, and the correct one for a floating value is 2 -- not 0, 1, or 3. On the implementation level, what happens now when you do not supply an integer, is that non-plural call is simply passed down to non-plural gettext() from libc. It is a fluke that for English it silently returns the plural, and not the singular string. For all I know, if the libc were built in develop/debug mode (I use distribution package), it could do something else. Therefore two independent KLocalizedString parameters are indeed necessary for the API, if you want to keep the integer/real alternation. Since it is about graph plotting, I suggest that you make the API such that the real- value string is mandatory, and the integer-value string optional. I.e. let the client choose if he wants alternation or not. Thank you very much your explanation - it makes a lot more sense to me now :-) Furthermore, I would suggest that the axis tick labels be pure numbers, and the axis name and unit shown separate in one place, e.g: -- 0 1 2 3 4 5 6 time [seconds] As an published engineer myself, I fully understand. But there's a few practical problems - we do not have much vertical space, for example. Putting on such an axis would pretty much halve the height available for the graph itself. If you were thinking of this for the vertical axis, then you are either asking for vertical text, which is evil, or a huge amount of wasted on the left hand side. John
