wish i++, i+=, i--, --i, ++i, i&= etc..

ivan.tellez · Aug 10, 2014

Erel said:
Have you seen any performance issues? In almost all cases the performance is the same as (well written) Java apps.

Maybe its not really a real performance issue, but I was inspired by:

canalrun said:
I remember a long time ago that the syntax, like +=, was done because compilers at the time could be made more efficient.

So i did some tests. I wrote some code using sums and increments, first in a Java Lib, then in B4A, the results were really conclusive. Java code in the lib was faster than B4A (tested in large loops).
For example a simple code in java:

B4X:

int z;
z++;

And then tested in B4A code

B4X:

Dim z as Int
z = z + 1

Code in java was 60% Faster.
Then just for curiosity tested

B4X:

Dim z as Int
z = z + "1"

Code in java was 37,000% Faster.

So, maybe could be more room for a new set of asigment operations with a more specific and optimized behavior (for example avoiding the implicit type convertion, only working if the operands are the same data type.

Picena Informatica · Aug 10, 2014

I think, with the market speed we have nowdays, that the more important thing is time to market. Other questions are relative. With B4A we have a lightspeed tool.

keirS · Aug 10, 2014

Erel said:
Have you seen any performance issues? In almost all cases the performance is the same as (well written) Java apps.

Most performance issues I have had are related to processing on the main thread. Most of those are to do with complex SQL queries. I have solved these by creating my own SQL library which gives me the ability to run queries on different threads. When I have finished writing a lazy loading cursor class I shall be releasing another version of my SQLExtended library with these features.

LucaMs · Aug 10, 2014

keirS said:
Most performance issues I have had are related to processing on the main thread. Most of those are to do with complex SQL queries. I have solved these by creating my own SQL library which gives me the ability to run queries on different threads. When I have finished writing a lazy loading cursor class I shall be releasing another version of my SQLExtended library with these features.

Are you sure you run the same query in the main thread directly in Java and have obtained higher response times?

I do not very much.

Anyway, this thread annoys me, "Unwatch" is a good solution

Erel · Aug 10, 2014

keirS said:
Most performance issues I have had are related to processing on the main thread. Most of those are to do with complex SQL queries. I have solved these by creating my own SQL library which gives me the ability to run queries on different threads. When I have finished writing a lazy loading cursor class I shall be releasing another version of my SQLExtended library with these features.

The SQL library supports asynchronous execution of commands and queries. See Asynchronous queries section here: http://www.b4x.com/android/forum/threads/sql-tutorial.6736/#content

Erel · Aug 10, 2014

Code in java was 37,000% Faster.

Can you post the java code? If you are not parsing the string like your B4A code does then your comparison is wrong.

Beja · Aug 10, 2014

-1

C & J people want to enter from the window huh! No way..
Next time someone will have a wish to eliminate the +1 altogether.

folks! forget ++

keirS · Aug 11, 2014

Erel said:
The SQL library supports asynchronous execution of commands and queries. See Asynchronous queries section here: http://www.b4x.com/android/forum/threads/sql-tutorial.6736/#content

Yes I know. For me the problem with ExecNonQueryBatch is that it doesn't give any feedback. I need to know the row id of inserted records and the count for updates and deletes. I would have to call ExecNonQueryBatch with one statement and then execute a query with either the changes() function or the last_insert_rowid() function. I think that would be rather counter productive and take more time than running everything on the main thread. So I wrote my own which gives me the feedback I need.

MaFu · Aug 11, 2014

Robert Valentino said:
I like this old animal called KEdit (came from the IBM mainframe world) and there are things I can do quicker in it that I cannot even do in other editors.

me too, KEdit rocks

ivan.tellez · Aug 12, 2014

Erel said:
Can you post the java code? If you are not parsing the string like your B4A code does then your comparison is wrong.

Hi, Im very sorry if i didnt explain myself. The 37,000% "Faster" java code is using only INTs without parsing a string. I know this is not a "fair" comparisson, but that was the point of the wish, adding the operators to have a chance to make them behave "different" than the current ones (for example not auto casting types and maybe more optimized).

As you said, "Adding features to a programming language should be done very carefully as it is a one way direction. You cannot later remove them."

Current operator cant change, but new ones (+=, -=, etc) could add support for programming some more robust app without deppending on lib creation.

By the way, testing the "fair" java code (parsing an string) , java code its 3x times faster (tested in large loops to se the difference):

B4A Code

B4X:

z = z + "1"

Java Code

B4X:

foo = Integer.parseInt("1");
z = z + foo;

Thraka · Aug 12, 2014

At minimum I would rather have += That still saves a lot of time and isn't just for a single thing like ++ is. ++ only gives you an increment and there is nothing else you can do with it. += allows you to put in any value, much more useful. (same for -=)

Nerdworld · Aug 17, 2014

As i remember, some other basic languages (DarkBasic?) at least had an "Increment"/"Decrement"-method, like:

B4X:

Dim a As Int = 5
Inc a ' 6
Inc a ' 7
Dec a ' 6
Dec a ' 5

yazak · Aug 18, 2014

Another basic I use Pure basic does it nice and simple

long form
a=a+1
short form is
a+1

Short form is faster as just adds 1 to a memory location so is faster
and Long form needs to load a add 1 then save a back to memory location.

But as Pure basic creates fast executable's its useful for speed in loops
but as for B4A unless there is a speed advantage to be made in java with shorter form may as well stick with a=a+1.

sorex · Aug 18, 2014

I think the inc/a+1 is only usefull speedwise if you compile it straight to assembler which is not the case with java, it's tokenized stuff.

still, abbrevations like this are always handy.

Troberg · Aug 21, 2014

++x and x++ also have some situations where it's not entirely obvious what will happen. Example:

a+++b

Is that a plus a pre-incremented b or is it post-incremented a plus b?

I don't actually mind the ++ notation, but it has some drawbacks.

Another nice shortcut for the most common use is the += operator, such as x+=1 to increment x by 1. Less likely to cause confusion, and more flexible.

wish i++, i+=, i--, --i, ++i, i&= etc..

ivan.tellez

Active Member

Picena Informatica

Active Member

keirS

Well-Known Member

LucaMs

Expert

Erel

B4X founder

Erel

B4X founder

Beja

Expert

keirS

Well-Known Member

MaFu

Well-Known Member

ivan.tellez

Active Member

Thraka

Member

Nerdworld

Member

yazak

Member

sorex

Expert

Troberg

Well-Known Member

wish i++, i+=, i--, --i, ++i, i&= etc..

Active Member

Active Member

Well-Known Member

Expert

B4X founder

B4X founder

Expert

Well-Known Member

Well-Known Member

Active Member

Member

Member

Member

Expert

Well-Known Member

Privacy & Transparency

Privacy & Transparency