1. [Phix] ? operator; string behavior

In Phix, the ability of question_mark ? being able to output text and numbers is really handy:

Statement Phix output OE output
? 'A' 65 65
? 'A' + 32 97 97
? 'A' & 'a' "Aa" {65,97}
? 65 & 97 "Aa" {65,97}
? 65 65 65
? {65} {65} {65}
? "A" "A" {65}
? "A" + 32 "a" 97
? {65} + 32 {97} {97}
? 65 + 32 97 97
? 65 & 97 "Aa" {65,97}

In Phix, single_quote ' delimiters sometimes make a number and sometimes a character. Any sometimes syntax is disturbing.

Phix

 
object x = { 'h','e','l','l','o' } 
? x 
puts(1, x) 
printf(1, "%s", {x} ) 
 
/* 
{104,101,108,108,111} 
ello?helloPress any key... 
*/ 
  • The 'h' is not output using the ? operator.
  • And, where does the extra '?' come from?

_tom

new topic     » topic index » view message » categorize

2. Re: [Phix] ? operator; string behavior

You found a bug! (actually in puts not ?, btw)

In builtins/VM/pfileioN.e, line 5195:

    [ELF64] 
--%rax  System call             %rdi                    %rsi                            %rdx                    %rcx                    %r8                     %r9 
--1     sys_write               unsigned int fd         const char *buf                 size_t count 
        mov rax,1               -- sys_write 
--      rdi already set         -- fd 
--31/10/17 (thanks to tom for finding this) 
--      mov rcx,rsp             -- buffer (==rdx btw) 
        mov rsi,rsp             -- buffer (==rdx btw) 
        mov rdx,1               -- count 
        syscall 

Despite having copied the api out for reference, I completely ignored it and set rcx instead of rsi, which just happens to be always pointing at the next character.

Obviously, after that 2-character edit, just run ".\p -c p" and it should be fixed.

Thanks, Pete

new topic     » goto parent     » topic index » view message » categorize

3. Re: [Phix] ? operator; string behavior

Testing Phix and WEE.

Irv's patch now makes WEE work in utf8: WEE, puts, terminal, office, and internet now all work in utf8.

Pete's Phix patch makes ? operator behave.

Phix ? operator

  • ? string of plain-text characters is output as text
  • ? string of utf8 characters is output as numbers
? "hello"        -- PLAIN TEXT  
    --> "hello" 
   
     
?  "ΕίČ░☼□"   -- UTF8 TEXT 
    -->{195,133,195,159,196,140,226,150,145,226,152,188,226,150,161} 
  • puts will output utf8 text
puts(1,"ΕίČ░☼□") 
--> ΕίČ░☼□ 
  • ? will output plain-text in its literal form
  • puts outputs text; control characters are active
-- literal (control characters not active) 
? "--- \n ---" 
 
/* 
"--- \n" ---" 
*/ 
-- control characters active 
 
puts(1, "--- \n ---" )     
/* 
--- 
--- 
*/ 

I now have WEE able to output utf8 to a terminal.

The Phix ? operator does not output utf8 text to a terminal.

Should the ? operator output utf8 strings?

_tom

new topic     » goto parent     » topic index » view message » categorize

4. Re: [Phix] ? operator; string behavior

_tom said...

Phix ? operator

  • ? string of plain-text characters is output as text
  • ? string of utf8 characters is output as numbers

There is a c>#7E on line 31 in function allascii in builtins\VM\psprintN.e that you might want to change to c>#FF

_tom said...

Should the ? operator output utf8 strings?

Can't say that has ever occurred to me. I generally view ? as a quick-and-dirty diagnostic that usually gets "it" (/the mind-reading thing) wrong anyway - eg this week I had a {3,4} that was not so helpful, so I whipped up a quick routine to show it as {PRE_SIB,POST_SIB}, somewhat more meaningful to me at the time. I might estimate that about 1 in 5 times I immediately replace a ? with pp(), printf(), or whatever, so I guess I have become pretty immune to any shortcomings. Another classic for me is ?ch --> printf(1,"%c\n",ch) or perhaps the ?""&ch or ?ch&"" trick.

No matter, the above change should do you proud anyways, since (I think) utf8 deliberately avoids bytes <#32.

Pete

new topic     » goto parent     » topic index » view message » categorize

5. Re: [Phix] ? operator; string behavior

The automagic is in the Phix 'print'

print(1, "cat") 
--> "cat" 
puts(1,"ΕίČ░☼□" ) 
--> "ΕίČ░☼□"  
print(1, "ΕίČ░☼□" ) 
--> {195,133,195,159,196,140,226,150,145,226,152,188,226,150,161} 

This makes Phix 'print' fundamentally different from OE 'print'.

Maybe Phix 'print' should have been named 'write' (because it is automagic).

_tom

new topic     » goto parent     » topic index » view message » categorize

6. Re: [Phix] ? operator; string behavior

_tom said...
print(1, "ΕίČ░☼□" ) 
--> {195,133,195,159,196,140,226,150,145,226,152,188,226,150,161} 

With the above change in place it matches the puts.

_tom said...

This makes Phix 'print' fundamentally different from OE 'print'.

Well, yeah, OE does not have a string type... They have always been different and always will be.

However, the documentation of print() does need updating to reflect that.

new topic     » goto parent     » topic index » view message » categorize

Search



Quick Links

User menu

Not signed in.

Misc Menu