Implement conio cgets #2705
Implement conio cgets #2705
Conversation
af3d0c4 to
17a8c92
Compare
|
Have you tested this on some other target(s)? Also, please add the function in the documentation That said, without testing, the function feels "too simple" to me - it should perhaps filter more (all?) "non printable" codes - else you get surprises on some targets, like eg when pressing the cursor keys on a C64 |
|
BTW
that should not be necessary, cgetc() should disable the cursor when needed, there shouldn't be any need to clean up behind it. To move the cursor left, without messing up anything else on the screen, a "backspace" should probably be converted into something like "cursor left, space, cursor left" - that would clean up the last character of course, not the cursor itself (perhaps that is what you mean?) |
|
Another thing.... i see some magic with CRLF - imho this shouldnt be required, cgetc() should always provide "\n" on "enter" (ie CH_ENTER) |
Can use ctype |
Yes, I mean the cursor itself. It "should" clear the cursor automatically prior to any conio The version without cursor clearing does work in C64. I don't have hardware or an original ROM for Apple][ so I can't test there. The options are:
In all cases, file a bug report in CX16: "conio goto functions not hiding cursor." |
Please don't "fix" it in cgets - it only hides an error that is elsewhere, and adds unnecessary side effects to the function itself for all targets. So please make sure the function is clean and works on a different target (c64,apple2,atari800 should be all mature enough for testing) and if it does, go ahead and try fixing the actual problem in the cx16 library :) |
In various sources, they use
Given that Ctrl-M for |
well, not really (using conio anyway) BUT that raises another question: do either '\n' or '\r' get added into the resulting string? IMHO - no :) |
include/conio.h
Outdated
| ** - terminating \0 is appended | ||
| ** - therefore the maximum number of characters which can be read is the size | ||
| ** of the buffer - 3! | ||
| ** |
There was a problem hiding this comment.
I see this just now... where does this "standard interface" come from? regular gets() doesn't work that way at least?
For conio, why not implement something like
// b points to buffer
// len is max len of the buffer (including the terminating 0)
// if len is 0, then len is not checked (to behave like sgets()
char *sgetsn(char *b, int len)
{
...
}
and then on top of that something like
char *cgets(char *b)
{
return cgetsn(b, 0);
}
There was a problem hiding this comment.
The "standard" interface I used is what I found online, some examples:
https://doxygen.reactos.org/df/db9/cgets_8c_source.html
https://www.pcjs.org/documents/books/mspl13/c/clibref/
https://github.com/gandrewstone/GameMaker/blob/master/tools/BORLANDC/CRTL/CLIB/CGETS.C
https://fossies.org/linux/TurboC/cgets.c
https://code-reference.com/c/conio.h/cgets
https://learn.microsoft.com/en-us/cpp/c-runtime-library/cgets-cgetws?view=msvc-170
https://www.ijaeast.com/C%20Logic%20Programming.pdf
The code layout you're suggesting with oddball cgets calling a saner version is what I'm implementing. I'm calling it cgetsx, +x for eXtra. If it's okay can make it public, too, at some later date, but would like to add cursor on/off, various echo settings, and CR/LF options, all would be specified as bit flags in a single unsigned char or int.
It turns out the cgets in the first draft is still too advanced; had to dumb it down even more; committed my latest. Currently trying to find Apple ][ and Atari 800 targets to test on. Works in C64 and handles multiple lines which I think is something not doable using fgets in that platform.
As mentioned before, half works in CX16, but leaves cursor remnants when backspacing.
don't overdo it - cgets should be just that, like gets but use the conio layer. If you want to do a more advanced general input function, i suggest doing that seperately. That said, passing parameters in the first two bytes of the buffer is just such a weird hack... please don't :) I would have never expected this in the borland api either... wth |
|
Agreed, I felt dirty writing it. But if it's not a "compliant" |
well, honest, conio is only losely based on the borland api. I would be very surprised if cgets() did work differentl to gets() |
|
If we're okay diverging from "standard" conio |
|
I wasn't able to find a documentation verification utility for SGML so hoping it's okay. Remaining is testing in Apple ][ and Atari 800 targets, in progress. |
|
looks ok to me now... one thing though: is that specific handling of wraparound at the end of a line really needed? I'd think that it is not shrug |
|
It would be great if I could just use I also tried Also can't do In summary, I think the line wrapping logic is required, and it's actually done more cleanly than some other implementations I've seen (excluding those where |
|
You are right... if course it doesnt work. I was too much thinking in terms of stdio. Guess i can merge then, thanks! |
|
Thank you. Much appreciated. |
The function
cgetsreads a string directly from the console. Usescgetcto retrieve character input, and other conio functions to handle multiline input and backspace. Like other conio functions it does not scroll the screen or check for writing out of bounds.Tested okay on CX16 emulator. It doesn't use any "tricks", other than when processing backspaces it outputs an extra blank just to make sure any cursor remnants aren't left behind*, so should be reasonably compatible with other platforms.
* reason is CX16 does have
cursorbut doesn't seem to honor the setting 100%!