Thread: About using global vars

In smaller projects, you shouldn't pass all-local variables. But in greater ones, you'll really mess things up. Look at namespaces. They're a great idea, but not used in C. Try organizing global structs.

I think we need examples. Post some common examples like "here a lot use a global, but it could be avoided". Some users, I suspect many, don't have the experience in working for a big C project so all of this seems kind of undefined. I have only made 2 medium-small C projects (3-4K lines) and didn't find the need to use globals at all. I want to do something, I create a function that takes input and gives output. The data I pass all in a struct which I update and flags are again all in one struct that I pass (or could just have had one global variable, no big difference).

Originally Posted by laserlight

The claims have been substantiated. You merely refuse to face the truth. "All you can give a lot of waffle basically."

In a way, you have a strawman argument, since it is indeed


Is it maintainable? Do you have a test suite? Are its components reusable with (almost) no modification?

It has been maintained, it started off basic and more and more functionality has been
added to it. It has also been 'ported' to work on 3 other poker sites, which store the
data in different ways, some put one history in a file, some several, some spread it over serveral folders, some use a mixture of the two, all use a different format for the history
with some similarities. It would prove too complex to have one multipurpose program
especially as I will add other sites of unknown formats.

Obviously I could have done some thing better with hindsight but not that much,
obviously you have to balance off forward planning with having a working version.

Each new feaure is tested.
It has been ran over a million poker hand histories, that is over a million files
in some cases. (obviously that take a while!!). It incorporates a database of a summary
of the previous results and loads it in real time and adds the next history to it.
You can get a commercial version called Pokertracker($98), I bet my version is quicker than that!! Obviously mine is not all singing all dancing but then it did not envolve
a million man hours!!

Originally Posted by laserlight

Clearly, this means that you lack the skill to participate in the development of programs of significant size.

You have provided no jusification for that statement.
If you mean projects which end up having to be scrapped then probably yes.
I have worked on a project with a heavy empathis on variable scope which had to be
rewritted from scratch for other reasons. So that was largely a waste of time.

Originally Posted by laserlight

No, the reason it is not a million lines long is because it is trivial.

It's not trivial just written by someone who is an efficient worker and knows how to
design and write a program which is actually useful.

Originally Posted by esbo

If you mean projects which end up having to be scrapped then probably yes.
I have worked on a project with a heavy empathis on variable scope which had to be
rewritted from scratch for other reasons. So that was largely a waste of time.

(snip)

It's not trivial just written by someone who is an efficient worker and knows how to
design and write a program which is actually useful.

Well don't talk as though local variables are the reason the projects that you work on tend to get scrapped. It is far more likely that you FUBAR'd them, if you had a significant role in the development of said software.

Also, Texas Holdem can be implemented trivially: you can write most card games for schoolwork if the I/O is lame enough, and you use rand as the PRNG. That's not a very convincing example on its own. Unless you want us to just give you the benefit of the doubt of course.

Perhaps that's for the best. You often worry about us stealing the code, maybe we'll steal the specs too!

Originally Posted by esbo

I don't think you pass a structure though do you?
You pass a pointer to a structure, so the function can modify the data.

The situation you described does not make the pointer any less local to a function.

Again it is hard to generalise about programs without a specific example.

Well apparently you fail to realize that if the shoe (the card decks in the game) and the shuffling algorithm are separate, self-contained entities it is trivial to integrate that code into another card game project, say, FreeCell. By spewing globals all over the place, you learn to rely on those and end up introducing interesting side effects to a program. That doesn't fascilitate code reuse. It isn't as easy as it could be to compile an already available shuffle, so you have to resolve the same, boring part of the project over and over again.

There are uses for global scoped stuff. But they are normally environment variables and not program data. Your inability to realise this is fascinating.

But then, you claimed to have maintained things before so who knows what sort of kluges you have up your sleeve to avoid whittling away at tedium.

You can guarantee he will usually hold AA KK or QQ, probably best to fold anyway

Every hand's a winner, every hand's a loser. You're playing with predictable people.

Originally Posted by whiteflags

Well don't talk as though local variables are the reason the projects that you work on tend to get scrapped. It is far more likely that you FUBAR'd them, if you had a significant role in the development of said software.

Nowt to do with me
Well maybe I spotted a big flaw in the design
I joined it midway through. (then everything went pear shaped - lol)

Originally Posted by whiteflags

Also, Texas Holdem can be implemented trivially: you can write most card games for schoolwork if the I/O is lame enough, and you use rand as the PRNG. That's not a very convincing example on its own. Unless you want us to just give you the benefit of the doubt of course.

What the program does is take a file such as below

Code:

Game #6682418603: Festive Freeroll (ID8701531) £0+£0 - Hold'em NL (75/150) - 2008/12/16 - 18:46:47 (UK)
Table "8701531 - 120" Seat 1 is the button. 
Seat 1: Jacq5050 sits out
Seat 2: Helokitty (2220 in chips)
Seat 3: esbo(1635 in chips)
Seat 4: _angela79 (4255 in chips)
Seat 5: Sveto2609 sits out
Seat 6: John135 (2630 in chips)
Seat 7: Arsim0205 sits out
Seat 8: Cojoc3101 sits out
Seat 9: Donn20044 sits out
Seat 10: __j_t__ (3050 in chips)
Helokitty: posts small blind 75
Flastard: posts big blind 150
----- HOLE CARDS -----
dealt to esbo[7c Ts]
_angela79: calls 150
Sveto2609: folds
John135: calls 150
Arsim0205: folds
Cojoc3101: folds
Donn20044: folds
__j_t__: folds
Jacq5050: folds
Helokitty: calls 75
Flastard: checks
----- FLOP ----- [2d 2s 6c]
Helokitty: checks
Flastard: checks
_angela79: checks
John135: bets 300
Helokitty: folds
Flastard: folds
_angela79: folds
Returned uncalled bets 300 to John135
John135: doesn't show hand
John135 collected 600 from Main pot
----- SUMMARY -----
Total pot 600 Main pot 600 Rake 0
Board [2d 2s 6c]
Seat 1: Jacq5050 (button) folded before Flop (didn't bet)
Seat 2: Helokitty (small blind) folded on the Flop
Seat 3: esbo(big blind) folded on the Flop
Seat 4: _angela79 folded on the Flop
Seat 5: Sveto2609 folded before Flop (didn't bet)
Seat 6: John135 collected 600
Seat 7: Arsim0205 folded before Flop (didn't bet)
Seat 8: Cojoc3101 folded before Flop (didn't bet)
Seat 9: Donn20044 folded before Flop (didn't bet)
Seat 10: __j_t__ folded before Flop (didn't bet)

And produce statistics on betting pattens etc (except properly aligned)

Code:

                                  8764794 - 18.txt
Marty02       52.0X        -9810P      24G     13C      25G     13C     5.0W  60.0X8  15.5CX3    21%WON     1F   4%FW    17%P  17%F   0%T   4%R 033%SD Marty02 
__simpy__     18.2X         7510P      28G      8C      11G      2C     5.0W  33.3X8  11.1CX3    18%WON     3F  11%FW     7%P  18%F  11%T  11%R 050%SD __simpy__ 
Juan1908      33.3X         2071P       9G      7C      21G      7C     5.3W  50.0X8  11.9CX3    59%WON     4F  44%FW    44%P  33%F  22%T  11%R 044%SD Juan1908 
Polipid1       0.0X        -1720P      15G      2C      13G      0C     1.0W   0.0X8   5.9CX3     7%WON     0F   0%FW     7%P   0%F   0%T   0%R 050%SD Polipid1 
esbo_____     16.1X        40265P   84161G  13960C   78607G  12685C  7947.2W  15.6X8  16.4CX3     9%WON  4251F   5%FW     4%P   6%F   2%T   2%R 052%SD esbo_____
Rammy         50.0X         9818P      77G     54C      18G      9C    32.0W  38.5X8  66.7CX3    42%WON    22F  29%FW    21%P  19%F  13%T   8%R 045%SD Rammy 
Tyrant002     50.0X          -15P       2G      1C       2G      1C     2.0W   NaNX8   4.5CX3   100%WON     1F  50%FW    50%P   0%F   0%T   0%R 100%SD Tyrant002 
Pme007         0.0X         -680P       1G      0C      14G      0C     0.0W   0.0X8   0.0CX3     0%WON     0F   0%FW     0%P   0%F   0%T   0%R NaN%SD Pme007 
Gav280216      0.0X         -870P       1G      0C      10G      0C     0.0W   NaNX8   0.0CX3     0%WON     0F   0%FW     0%P   0%F   0%T   0%R NaN%SD Gav280216 
Poshtosh      46.2X        20230P      16G      7C      13G      6C     4.0W  46.2X8  21.9CX3    25%WON     3F  19%FW     6%P  12%F  19%T  19%R 050%SD Poshtosh

So " 5%FW 4%P 6%F 2%T 2%R 052%SD esbo_____" tells me I 'force' 5% of my wins, raise preflop 4% of the time, raise 6% of the time post flop, raise 2% of the time after the 'turn' and raise 2% of the time after the 'river' (last card), and win 52% of my 'showdowns'.

Other sites use different formats but some stuff is broadly the same.
That was a fairly simple game as not much bettiing too place.
Some sites miss information out so you have to work it out from the info given.

Also there are different types games, cash and tournements etc..

Originally Posted by whiteflags

Perhaps that's for the best. You often worry about us stealing the code, maybe we'll steal the specs too!


The situation you described does not make the pointer any less local to a function.


Well apparently you fail to realize that if the shoe (the card decks in the game) and the shuffling algorithm are separate, self-contained entities it is trivial to integrate that code into another card game project, say, FreeCell. By spewing globals all over the place, you learn to rely on those and end up introducing interesting side effects to a program. That doesn't fascilitate code reuse. It isn't as easy as it could be to compile an already available shuffle, so you have to resolve the same, boring part of the project over and over again.

There are uses for global scoped stuff. But they are normally environment variables and not program data. Your inability to realise this is fascinating.

But then, you claimed to have maintained things before so who knows what sort of kluges you have up your sleeve to avoid whittling away at tedium.


Every hand's a winner, every hand's a loser. You're playing with predictable people.

Wish I was I like predictable people!!
People get more predictable as the stakes get higher, apart from the 'nutters'

Originally Posted by laserlight

Note that I did not ask whether the program was reusable. I asked whether the components are reusable. Trying to write a program to solve every possible future requirement is indeed a fruitless task. Trying to write a program that can be extended with minimal changing of the current source code, and whose components can be reused in a new program with similiar yet different requirements is both feasible and desirable.

The 'components' as such are to specific to the program, that the chance needing to
reuse then is so small that making them reusable is a waste of time.
I sometines copy the odd bit out, for example opening a few files, not worth doing much for than that.

Originally Posted by laserlight

You have provided no jusification for that statement. A 1200 line C++ program is typically trivial, at least when compared to say, the complexity and resulting number of lines of code required by a modern operating system. Of course, even a program with less than 100 lines of simple code can be useful, so by saying that your program was trivial I did not imply that it was not useful.

Those operating systems are so big because they are badly designed.
It is because they fail to identify common components and constantly rewrite them,
they fail the test of reusablity. My programs are small because I don't keep re-inventing
the wheel.
The size of some programs these days is staggering beyond belief, what do they do?
Write every loop out!!?

I encounter some apallingly bad practices using programs, take this edit box for example.
It is a pathetic design.
For example, if I move the cursor past the end of this line and click, it goes to the end of the like - it's complete garbage. If I press down arrow it moves to the start of the next line, not just down - absolute rubbish.

Originally Posted by laserlight

That implies that your functions are not designed to be cohesive, loosely coupled units. They do not "do one thing and do it well". This in turn makes them monolithic components that are hard to reuse.

No it does not. Most of the functions are so specific that they could never concieveable
be reused in another program. They are basically just bits of code which the program and that program alone uses several times. Thats why it is not a billion lines long!!
Big programs tend to be bad programs.

Originally Posted by laserlight

This contradicts your own statement that "It has also been 'ported' to work on 3 other poker sites, which store the data in different ways, some put one history in a file, some several, some spread it over serveral folders, some use a mixture of the two, all use a different format for the history with some similarities."

There was one routine which I wish I had written a little better, it would have made it
easier to port and has been one of the things I have been meaning to do for a while
but as it does make much difference anyway I have not bothered yet.
Just a simple thing really, I should have 'tied' the length of a string to the striing.
eg strptr=inputstring, strnglen=strlen(inputstring);
It's not much work to modify though, and it would not always 'work' anyway.
The stuff you mention above would have been too difficult to implement really
especialy as I don't know all the formats used, and I really needed a working version
for my main site, it would literally have cost me money to try and implement that.
It's easy to say in an ideal world where you have no time or cost constraints.

One puts it histories in reverse time order ie newest first rather than last.

Originally Posted by laserlight

Since you are unable to reuse your components, it is obvious that you are merely bluffing.

There are just not really any cmponents which can be reused.
C already has a vast library or reusable components.
Most of the time it is qucker to write your own routine thann search the librarys!!

If you think you can extract all the data in less lines you are welcome to try!!