-
|
Is there a way I can use the CB for handling function pointers. I’m trying to use it to queue multiple events that are pushed when either a timed or a pin interrupt occurs. |
Beta Was this translation helpful? Give feedback.
Replies: 7 comments 16 replies
-
|
Yes, there’s no reason why you shouldn’t! You just declare a buffer of your type, which is going to be the function reference of your choice: typedef void (*func)();
CircularBuffer<func,100> functions;
void a() {}
void b() {}
void c() {}
void setup() {
functions.push(a);
functions.push(b);
functions.push(c);
}Please, note I haven’t compiled the code above, I just pulled it out while on the go, it’s just a reference to describe how you can do it. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your fast reply. I decided to add some additional features so I'm now using a class. Although the VS19 C++ editor does not show any errors, but the output shows some issues that I'm not sure how to "repair". I'm a C# developer who hasn't coded C++ in 25 years, so please forgive my ignorance. Here's the output. I'm creating a timed interrupt library for simple interrupt handling. Definitely not RTOS. NOTE-DefinedValues and Enums contain items that are related to my full project - A Maple Sap Pump/Vacuum collection system - both will be open source. |
Beta Was this translation helpful? Give feedback.
-
|
First of all, I had a quick look at your code and I don't think the problem is something in the CIrcularBuffer library, but in your code. For what I can understand from your code, you are trying to wrap a When instantiated in this way, the Now, as a Java programmer myself I'm having difficulty in understanding the logic behind your code as it seems you are mixing up responsibility between the Hoping this solves your doubts. |
Beta Was this translation helpful? Give feedback.
-
|
I did not think it was your code, but my implementation. I'll be trying to restructure my solution. |
Beta Was this translation helpful? Give feedback.
-
|
I got everything working except I seem to be screwing up memory. For the new classes, do I have to do a malloc, or something like that. I've eliminated the InteruptBuffer and am able to push 5 objects into the buffer. When I try to pop them, the first pops properly, the rest are ??? I know it's not the CB. I'm just not used to c++ and am not sure of how to proceed. |
Beta Was this translation helpful? Give feedback.
-
|
I don't see any reference to this library in the code you have attached: I believe you'll get better and faster help on StackOverflow, here we aim at providing support to the library users. Apologies for being unable to provide additional information, but you are asking for a generic C++ code review to someone who is not a C++ expert. |
Beta Was this translation helpful? Give feedback.
-
|
The reference is in interupt.h: Since I am one of the founders of Infragistics.com (1990), I have had to deal with others duplicating our product offerings, so I understand how you feel. I am just frustrated that my ignorance (lack of experience) with C++ has been causing me problems. C# has been my primary language. I will follow up with StackExchange. Once I get this working, I'll make it into a library (that requires your circular buffer). It will be a poor man's RTOS for the arduino, supporting only timed/pin interrupts but not time slicing. Sorry if you feel that I used your time poorly. When it's up and running I'll let you know. La ringrazio molto per il vostro tempo. Mia moglie ha vissuto a Bolongna per molti anni. Ma non parlo molto bene l'italiano. Abbott |
Beta Was this translation helpful? Give feedback.
-
|
I meant StackOverflow ... |
Beta Was this translation helpful? Give feedback.
-
|
No man, I might have been misleading: I have absolutely no problem with the creation of an alternative to this library and I'm not concerned with duplication, it's just I know how painful it can be to debug code on a microcontroller and also I'm truly not a C++ expert. If you are trying to put a series of functions in a CircularBuffer, here is the code I would use: #include <CircularBuffer.h>
// function signature
typedef uint8_t (*func_t)(void);
// the buffer
CircularBuffer<func_t, 10> buffer;
void setup() {
Serial.begin(9600);
// just an initial fill up
buffer.push(fn1);
buffer.push(fn2);
buffer.push(fn3);
buffer.push(fn4);
}
void loop() {
if (!buffer.isEmpty()) {
// the use of push + shift realizes a FIFO (also unshift + pop would do)
func_t fn = buffer.shift();
// invokes the code in the function and prints the result code
Serial.println(fn());
}
}
uint8_t fn1() {
Serial.println("fn1");
return 0;
}
uint8_t fn2() {
Serial.println("fn2");
return 1;
}
uint8_t fn3() {
Serial.println("fn3");
return 2;
}
uint8_t fn4() {
Serial.println("fn4");
return 3;
}If, instead of functions, you want to push objects then there's an example in the library To what I can understand from the code you have submitted I'm not sure a circular buffer is what you are looking for: a CB implies the concept when too much stuff is added, some other is simply discarded... if you are going to use such a buffer to implement a RTOS, then when you try to enqueue more stuff than viable you either discard the most recently added or the oldest... Also, you are trying to add a priority concept... To me, it looks you are trying to create a priority queue based on a circular buffer, but such implementation would be much different because you'll have to either browse the queue by priority or populate it based on priority, meaning the backing structure cannot be an array (like the one I'm using), but rather a linked list or a double linked list (if you can trade memory for better performance). So, it's not a matter of time, but rather a much bigger topic... |
Beta Was this translation helpful? Give feedback.
-
|
I've had a second look at your InteruptorClass InteruptClass::runNextInterupt()
{
// the following instruction retrieves the first element, also known as *head*
// NOTE: you are not checking for buffer emptiness, meaning you might be retrieving invalid memory!
InteruptorClass* interuptor = InteruptClass::peek();
if(interuptor->canRunFunction()){
interuptor->runInterupt();
// the following instruction extracts from the tail, also referred as *last element*
InteruptClass::pop();
interuptor->deleteMe();
}
else
{
// do nothimg - where waiting for period to elapse
}
}As you can see, you are executing one interrupt but deleting another one... I would restructure the code like the following, eliminating all the additional wrapping methods: // define your own method to add interrupts
InteruptorClass InteruptClass::enqueue(void (*fPointer)(InteruptorClass interupt),
EnumsClass::Interupt fifoLifoPush,
EnumsClass::Interupt functionName,
unsigned long period)
{
cBuffer.unshift(new InteruptorClass(fPointer, fifoLifoPush, functionName, period)); // adds a new item to the head
}
// I would personally call this simply *next()* or *run()*
InteruptorClass InteruptClass::runNextInterupt()
{
if(!cBuffer.isEmpty() && cBuffer.last()->canRunFunction()) { // inspects the tail
InteruptorClass* interuptor = cBuffer.pop(); // extracts the tail
interuptor->runInterupt();
// I wouldn't define a method to wrap a call to delete, you decide how objects are created and thus destroyed
// so as I've used the *new* operator above I'm using the *delete* operator here
delete interuptor;
}
else
{
// if it does nothing then remove the block
}
}Your code is also hard to follow because of the names you have picked up: Finally, the code flow is totally unclear to me and, honestly, overcomplicated. |
Beta Was this translation helpful? Give feedback.
-
|
First of all - a Happy and a safe New Year. has compile error They all throw errors ??? enqueue defined as: Please let me know when I am annoying you too much. |
Beta Was this translation helpful? Give feedback.
-
The IHandlerClass::enqueue(testFunction, EnumsClass::FifoPush, EnumsClass::CheckInTemp, 2);I provided an example using a This thread is clearly no more related to the library or it's proper usage, looks like a support thread on C++ programming for which you can find better and more correct help than what you can get here. My last generic hint: having one enumeration listing everything is useless and not different from a series of constants: the whole concept behind enumerations is to define distinct sets, so you want to have an enum for enqueing behaviours and another one for task types. |
Beta Was this translation helpful? Give feedback.
-
|
My previous example corrected to your new names... // defines the function prototype, I'm not sure if you need a reference to the task
typedef void (*function_t)(Task task)
typedef struct {
unsigned long executed;
unsigned long period;
unsigned int repetitions; // how many times this task should be repeated, once every period
bool bubbleUp; // equivalent of you LifoPush/FifoPush: if true repeats as first
EnumsClass type; // I wonder if this is necessary, to me this might be implied by the function
function_t func;
} Task;
bool IHandlerClass::enqueue(function_t func, unsigned long period, unsigned int repetitions, EnumsClass type, bool bubbleUp) {
return cBuffer.unshift(Task{millis(), period, repetitions, type, bubbleUp, func}); // adds a new item to the head
}
bool IHandlerClass::runNextTask() {
if(!cBuffer.isEmpty() && this->canRunTask(cBuffer.last())) { // inspects the tail
Task task = cBuffer.pop(); // extracts the tail
task.func(task);
if (--task.repetitions > 0) { // decrements repetitions
task.bubbleUp ? cBuffer.push(task) : cBuffer.unshift(task); // re-adds the item to the tail or the head depending on the bubbleUp param
}
return true;
} else {
return false;
}
}
bool IHandlerClass::canRunTask(Task task) {
_currentMillis = millis();
if(millis() - task->executed >= task->period) {
task->executed = millis();
return true;
} else {
return false;
}
} |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for all your support. I was able to get everything going except for the function pointer ... which I'll follow-up on StackOverflow. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your follow-up. I am dealing with a dynamic array of class object, and all the samples I see are for statically created objects. Once I can get the dynamic class array working properly, I can then use your buffer properly. |
Beta Was this translation helpful? Give feedback.
-
|
The thing is on an MCU you have very little memory and also there is no real distinction between heap and stack as you don't have an underlying OS: in other words, you must have a very specific reason to use dynamically allocated objects as you are entering a very dangerous realm. So, I'm not telling you what to do, I'm just raising a yellow flag so that you know what you are going to have to deal with. |
Beta Was this translation helpful? Give feedback.
-
|
You're quite right. I am creating a static array of sufficient classes (I now call them tasks - thank you) to support the interrupt activity, As an interrupt occurs, I pick the first unused task. This way I will not commit the error of constantly allocating and de allocating memory. , InteruptClass uses the unused task ,sets its properties and pushes it onto your CB, including the function pointer. Once the task is complete, the task is reset for reuse. |
Beta Was this translation helpful? Give feedback.
-
|
So your I took the liberty to add a // defines the function prototype, I'm not sure if you need a reference to the task
typedef void (*function_t)(Task task)
typedef struct {
unsigned long executed;
unsigned long period;
unsigned int repetitions; // how many times this task should be repeated, once every period
function_t func;
} Task;
bool InteruptClass::enqueue(const function_t func, const unsigned long period, const unsigned int repetitions) {
return cBuffer.unshift(Task{millis(), period, repetitions, func}); // adds a new item to the head
}
bool InteruptClass::runNextTask() {
if(!cBuffer.isEmpty() && this->canRunTask(cBuffer.last())) { // inspects the tail
Task task = cBuffer.pop(); // extracts the tail
task.func(task);
if (--task.repetitions > 0) { // decrements repetitions
cBuffer.push(task); // re-adds the item to the tail, but you might want to add it to the head, depending on the behaviour you want to achieve
}
return true;
} else {
return false;
}
}
bool InteruptClass::canRunTask(Task task) {
_currentMillis = millis();
if(millis() - task->executed >= task->period) {
task->executed = millis();
return true;
} else {
return false;
}
}The great advantage of using a struct is the whole memory necessary to hold the circular buffer is allocated when the circular buffer is initialized and can be checked at compile time. |
Beta Was this translation helpful? Give feedback.
-
|
@AbbottHMG I did something similar, having a dynamically growing array in ESP8266's heap of class/structs. I used a modified version of |
Beta Was this translation helpful? Give feedback.
-
|
Thanks. I'm looking at it now. |
Beta Was this translation helpful? Give feedback.
-
|
It would be nice if you mark the answer which best responds to your question. Thank you |
Beta Was this translation helpful? Give feedback.
-
|
This is the working test version of my task manager. The task manger controls the CB and the task array that it points to. It also fills the CB with objects in the array. The ino loop empties the buffer. |
Beta Was this translation helpful? Give feedback.
-
|
I'm only asking you to select here the answer which best responds to your question. |
Beta Was this translation helpful? Give feedback.
Yes, there’s no reason why you shouldn’t! You just declare a buffer of your type, which is going to be the function reference of your choice:
Please, note I haven’t compiled the code above, I just pulled it out while on the go, it’s just a reference to describe how you can do it.
Also, by varying the typedef you can establish input parameters and return values: you are not limited to functions without parameters or return values, but they must all comply with the same interface.