Provide access to left-over milliseconds

[jamesmyatt]

These are some minimal changes to provide access to the left-over millis from a call to now(). I know there's been lots of discussion about this, but I need it for my clock animations and I've included suitable warnings in the Readme. Furthermore, you have to enable access to the functions via a #define in your sketch, which I think is a pretty good compromise.

I think this is better than the solutions in #11 and #45, since it will be consistent with the seconds output.

Datatype for the sysTimeMillis was selected to match prevMillis, which is uint32_t, rather than millis(), which is unsigned long.

[ramarro123]

any plan to merge it?

[jhdrn]

This is something I'm looking for too, so please merge this! :)

[cyberman54]

This would also meet my requirements.

[nobodyinperson]

I'd rather do this in Time.h instead of hard-coding it in the compiled .cpp-file and test via #if TIMELIB_ENABLE_MILLIS:

#ifndef TIMELIB_ENABLE_MILLIS
#define TIMELIB_ENABLE_MILLIS true
#endif // #ifndef TIMELIB_ENABLE_MILLIS

This way, it is possible to pass -DTIMELIB_ENABLE_MILLIS=false to the compiler without getting an annoying duplicate-definition warning.

[jamesmyatt]

I'm not sure this would work since the later code in Time.cpp relies on all of the declarations in Time.h. I think it would be better to #undef TIMELIB_ENABLE_MILLIS if it's already defined at this point.

[nobodyinperson]

It will work, I am doing it in all my projects. The problem is that the .cpp file is compiled in a different translation unit, so to change compiler flags inside it you have to either change it manually in the code (yuck!) or provide a global compiler flag. AFAIK it is not possible to tell the compiler to undo a specific #define, so it is effectively impossible to toggle TIMELIB_ENABLE_MILLIS from outside the code.

Solutions are:

1. removing any #define TIMELIB_ENABLE_MILLIS from any files, keeping the #ifdef checks and letting the user provide a command-line compiler flag. This however is inconvenient (if not impossible) for the Arduino IDE.
2. my suggestion: provide a boolean default in the header file (which is #included in the source code file, thus causes a rebuild) and change the checks to #if. This way, a global compiler flag can be used to explicitly enable or disable the feature (vs only being able to enable it as in solution 1). Still, the Arduino IDE problem.
3. Making this whole library a header-only library and using my suggestion (solution 2). This would enable users of the library to simply #define TIMELIB_ENABLE_MILLIS true/false before #includeing the library to toggle the feature. No need for a global compiler flag, no problem in the IDE. That's what I am mostly doing. Has its own disadvantages. But that would be up to @PaulStoffregen to decide.

[jamesmyatt]

I'm all for avoiding the warning. However, I don't see the problem of forcing TIMELIB_ENABLE_MILLIS to be set in Time.cpp. As you say it's a separate translation unit, so it doesn't change the functions available to the user when they include TimeLib.h.

The purpose of this flag is to enable the declarations involving millis in TimeLib.h so that an advanced user can say "I really want access to the leftover millis. I understand the risks and limitations. I promise I know what I'm doing and I won't complain". So with this solution, you can already just #define TIMELIB_ENABLE_MILLIS in your sketch immediately before #include<TimeLib.h> in your source code (this is what I do). No need for extra compiler flags.

Also I don't see why you would ever want to undefine it. Either you do nothing and the millis aren't available in your sketch, or you use the millis in your sketch so you define it in the sketch. Am I missing something?

[nobodyinperson]

I see. You're right.

Funny way of using compiler flags to enable features. I guess, you could ditch all checks for TIMELIB_ENABLE_MILLIS in the .cpp-file altogether then, right? Can't undefine it via compiler command-line flags, hard-coded in the implementation file, feature is only enabled in the header file. So the milliseconds support is always compiled, right?

I guess one could leave it this way. Still, I prefer being explicit than implicit (i.e. being able to explicitly en-/disable some feature). But maybe that's just me. Still up to @PaulStoffregen to merge this into his library.

[lkaupp]

Thank you very much for this patch, merged manually, works like a charm!

[cyberman54]

any chance to get this PR into the main release?

[cyberman54]

any news / plans here?

[jamesmyatt]

@cyberman54 , I think it's unlikely that this will get merged

[jamesmyatt]

For anyone watching, it turns out the functions in the ESP8266 and ESP32 frameworks have functionality like this, e.g. gettimeofday.

[karelv]

👍 I want it!