which is the best way to declare Serial while creating Arduino library?

Question

Well, I am trying make a library for HMI display. display and Arduino comminicate with Serial. So I will so many times Serial port. Generally I will use Serial.write commend. if I examine other libraries, there are lots way to declare serial. Why and which is the best I want to learn.

there I have seen declaring Serial methods.

Method 1

in library .h file

nothing

in library .cpp file

#include <SoftwareSerial.h>
SoftwareSerial SIM(RX_PIN,TX_PIN);
void KapadokyaGSM::basla(){
 SIM.begin(9600);
 _tampon.reserve(255);
}

Method 2

in library .h file

public:
EasyNex(HardwareSerial& serial);
...
private:
HardwareSerial* _serial;
...

in library .cpp file

EasyNex::EasyNex(HardwareSerial& serial){ // Constructor's parameter is the Serial we want to use
_serial = &serial;
}
void EasyNex::begin(unsigned long baud){
_serial->begin(baud); // We pass the initialization data to the objects (baud rate) default: 9600
delay(100); // Wait for the Serial to initialize

Also I have seen declaring with Stream but now i didnt find example codes. So finally Can anyone tell me what is the best for declearing Serial in library. it can be for only hardware serial or both

• chrylis -cautiouslyoptimistic-: The basic concept that you're asking about is called dependency injection.

Answer 1

As regards the use of Stream:

I usually use Stream instead of HardwareSerial because:

• It allows the use of other serial devices, like SoftwareSerial or USBSerial that aren't "Hardware" UARTs.
• It even allows the use of non-serial systems, like sending data over networks or wireless devices

However it means:

• The user has to manually initialize their serial object / device as Stream has no facility for setting things like baud rates.

• mehmet: Can you show me a tutorial, or github library which is using Stream.
• Majenko: @mehmet try this one: github.com/MajenkoLibraries/ICSC/tree/master/src

Answer 2

There is no best way, the first is using a global variable, being static. The second is dynamically (or at least to pass a serial instance).

It depends on your use

• If you want your code always to work with the same declared serial (hardware or serial) instance, you can use the first method, as it is the easiest.
• If you want your serial hardware port to be flexible (your code to work e.g. with a selectable hardware serial port on an Arduino Mega), you need the second.
• If you want to your serial port to be either software or hardware, you can use the second method, but you need another type (not the mentioned HardwareSerial class. I don't know by head the parent class of SoftwareSerial and HardwareSerial (The Stream class you mentioned could be a common parent class).

• Juraj: there is no Serial base class
• mehmet: Can i ask a last question, why second method uses "->" and Why not "." Such us _serial.begin(baud). Is this totally same
• JRobert: If you have access to the object itself, e.g., Serial, you can say Serial.println(). If your library only has a pointer to the object, you have use ->, the dereference operator: serialPtr->println(). The result is the same: calling the object's 'println()' member. The difference is in how the compiled code gets you to that member.
• Michel Keijzers: @Juraj Thanks for that addition (I was not sure, so I removed the comment about not being sure and the statement in my answer).
• Juraj: @mehmet, it is common C++ object member access you asked about