digitalmars.D - ConvExt.d v0.2 please review, comment, suggest, and / or correct.
- David L. Davis (520/520) Apr 09 2005 Well per Walter advice in to below message,
- Andrew Fedoniouk (11/575) Apr 09 2005 Little bit out of scope but I would like to rise question about
- Ben Hinkle (40/223) Apr 10 2005 cool. comments inline
- David L. Davis (49/174) Apr 10 2005 Yeah, the comment style is what I started to use on code on my web-site
- Ben Hinkle (11/48) Apr 11 2005 Your chances of getting something done to phobos goes up if you make it ...
Well per Walter advice in to below message, D forum thread: "process improvement for phobos" http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/20328 I'd like to resubmit to my fellow D developers for comments and review, the following code that's been submitted to Walter for including into Phobos. Recently, D v0.120 has forced a number of correction to be made into this code (a lot of additional cast()s had to put into place...which was kool, a better compiler means better executing code), but in so doing I found some areas that needed a bit more tighting up. Please free feel to give some suggests and / or correction to the following code, that's designed to fill in the missing "string to floating-point" function gap in std.conv. Note I'm in no way a floating-point expert, but just a guy trying in improve and move D forward...one line or one function at a time if I possibly can. Thanks in advance, David L. ------------------------------------------------------------------- "Dare to reach for the Stars...Dare to Dream, Build, and Achieve!"
Apr 09 2005
"Andrew Fedoniouk" <news terrainformatica.com> Little bit out of scope but I would like to rise question about string-to-value function signatures again. My proposal is to use int valueOf( string s , out T v, in T defaultValue = 0 ) Where T is a numeric type int, uint, etc. Functions return number of valid characters processed. Such functions are flexible enough to allow blind and controllable conversions. Andrew. "David L. Davis" <SpottedTiger yahoo.com> wrote in message news:d3903a$vr8$1 digitaldaemon.com...Well per Walter advice in to below message, D forum thread: "process improvement for phobos" http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/20328 I'd like to resubmit to my fellow D developers for comments and review, the following code that's been submitted to Walter for including into Phobos. Recently, D v0.120 has forced a number of correction to be made into this code (a lot of additional cast()s had to put into place...which was kool, a better compiler means better executing code), but in so doing I found some areas that needed a bit more tighting up. Please free feel to give some suggests and / or correction to the following code, that's designed to fill in the missing "string to floating-point" function gap in std.conv. Note I'm in no way a floating-point expert, but just a guy trying in improve and move D forward...one line or one function at a time if I possibly can. code runtime toIfloat(), which I've functions. ' -------------------------------------------------------------------------- function char[] ).unittest" ); char[] ).unittest" ); real.max ) ); char[] ).unittest" ); cast(real)ifloat.min.im ) ); 123.45i ) ); 456.77i ) ); char[] ).unittest" ); cast(real)idouble.min.re ) ); cast(real)idouble.min.im ) ); cast(real)idouble.max.re ) ); cast(real)idouble.max.im ) ); char[] ).unittest" ); ireal.min ) ); ireal.max ) ); char[] ).unittest" ); cfloat.min ) ); cfloat.max ) ); char[] ).unittest" ); char[] ).unittest" ); creal.min ) ); creal.max ) ); cast(real)creal.max.re ) ); cast(real)creal.max.im ) ); cast(real)1.2345e-5+0i.re ) ); cast(real)1.2345e-5+0i.im ) ); 0.0e-0+0i ) ); s2 ) Thanks in advance, David L. ------------------------------------------------------------------- "Dare to reach for the Stars...Dare to Dream, Build, and Achieve!"
Apr 09 2005
"David L. Davis" <SpottedTiger yahoo.com> wrote in message news:d3903a$vr8$1 digitaldaemon.com...Well per Walter advice in to below message, D forum thread: "process improvement for phobos" http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/20328 I'd like to resubmit to my fellow D developers for comments and review, the following code that's been submitted to Walter for including into Phobos. Recently, D v0.120 has forced a number of correction to be made into this code (a lot of additional cast()s had to put into place...which was kool, a better compiler means better executing code), but in so doing I found some areas that needed a bit more tighting up. Please free feel to give some suggests and / or correction to the following code, that's designed to fill in the missing "string to floating-point" function gap in std.conv. Note I'm in no way a floating-point expert, but just a guy trying in improve and move D forward...one line or one function at a time if I possibly can.cool. comments inlineMy first comment is the comment style. Does anything else in phobos use this comment style? Looking in std.conv I see comments like /************************* * Convert blah to foo. * Gammar: digit(digit) */ The use of /+ and ' seems non-standard. I'm not sure exactly what you intend to send Walter but if this stuff is to go in std.convext (though I don't see why it can't go into std.conv if you ask me) I would get it into the exact format that you want to appear in phobos. It looks like none of the functions have comments preceding them in the style of std.conv - those comments should be in there to explain just what is allowed and what isn't.code runtime toIfloat(), which I've functions. ' -------------------------------------------------------------------------- functionPersonally I'd remove these toBit conversions. The strings it recognizes seem somewhat arbitrary. Why "yes" but not "oui"? If anything I would just have toBit recognize what std.format does with a bit, which is print "true" or "false". And given that restriction it would be simpler to just have a comparison with "true" or "false". One trouble with toBit(char c) for example is the numeric nature char might lead people to guess that c!=0 is the test for toBit instead of comparing with some ascii codes.All these floating point functions seem to be very small wrappers around std.math2.atof. Why not put these in math2? I can see the desire to mimic the std.conv module but IMO that isn't a strong enough reason to make another module. Also if the idea is to follow std.conv then I would have the ones that cast check for overflow etc just like toShort checks for overflow after getting the result of toInt. [snip]why 0f?I don't see where the index of the "i" shows up. If you only parse "<float>i" then you can just check the last char instead of doing an ifind.I'm not sure what f.nan == true is supposed to test. Do you want to see if f is a nan? The isnan function in std.math is for that.[snip]s2 )why the dup? You shouldn't need to change s so there is no need to make a copy. Also I can't tell exactly what string parse into complex numbers. For example will the string "1-2i" parse correctly? I think it should. In any case documenting what works and what doesn't is important.
Apr 10 2005
In article <d3cq53$19hh$1 digitaldaemon.com>, Ben Hinkle says..."David L. Davis" <SpottedTiger yahoo.com> wrote in message news:d3903a$vr8$1 digitaldaemon.com...Yeah, the comment style is what I started to use on code on my web-site (something I remembere picking up from my programming classes back in my college days), but I don't really expect Walter to keep them. :) And since I've passed all rights of the code to him, he can make the changes he needs to make.My first comment is the comment style. Does anything else in phobos use this comment style? Looking in std.conv I see comments like /************************* * Convert blah to foo. * Gammar: digit(digit) */The use of /+ and ' seems non-standard. I'm not sure exactly what you intend to send Walter but if this stuff is to go in std.convext (though I don't see why it can't go into std.conv if you ask me) I would get it into the exact format that you want to appear in phobos. It looks like none of the functions have comments preceding them in the style of std.conv - those comments should be in there to explain just what is allowed and what isn't.One of the nice things about D is that it gives you choices...in this case two multi-commenting blocks like "/* */" and "/+ +/". But it's hard to find a color-code syntax within the text editor to support more than one multi-comment block set of symbols. So, I picked to setup the "/+ +/" within the editor. And personally I like it a lot more than the "/* */".The toBit() functions were a bit of me thinking out loud I guess, cause I really think that the std.conv module should cover all the datatypes. But here again, Walter will pick what he wants to add...and even then he will rewrite it if neccessary.Personally I'd remove these toBit conversions. The strings it recognizes seem somewhat arbitrary. Why "yes" but not "oui"? If anything I would just have toBit recognize what std.format does with a bit, which is print "true" or "false". And given that restriction it would be simpler to just have a comparison with "true" or "false". One trouble with toBit(char c) for example is the numeric nature char might lead people to guess that c!=0 is the test for toBit instead of comparing with some ascii codes.It true that these are mainly wrapper functions that rely heavily upon std.math2 module, but it just makes more sense to me that std.conv is the module that converts strings (char[]s) back to numbers (both whole numbers and floating-point). But I'm beginning understand more as to why you were suggesting in the fromString() functions within some comments a few weeks ago in a different thread. It would be a nice way to have the toString() mirrored with the fromString() functions which could be overloaded with whatever numercial datatype that were passed into them. I'll have to take another look at the toShort() and toInt() functions in regards to the overflow checking. Thanks for the heads up on that.All these floating point functions seem to be very small wrappers around std.math2.atof. Why not put these in math2? I can see the desire to mimic the std.conv module but IMO that isn't a strong enough reason to make another module. Also if the idea is to follow std.conv then I would have the ones that cast check for overflow etc just like toShort checks for overflow after getting the result of toInt.[snip]I thought it better to cast the numercial literal with a suffix ("f" for float). Is that the wrong way to do it? I could have used cast(float)0 instead, would this had been better?why 0f?You're right, the "i" would be at the end, if it were passed in. Of course the could "i" be missing from string all together, so instead of writing an if statement I used the ifind() function. Which on second thought, maybe I should've used the "if" statement. Thanks for pointing it out.I don't see where the index of the "i" shows up. If you only parse "<float>i" then you can just check the last char instead of doing an ifind.I wanted to put in code to check for a nan value, but to tell the truth I may not totally understand the use behind them. I thinking that a NaN is much like a null, but only for floating-point. Also, I looked at using the isnan() function, but it can't handle the <float>i (the "i" part of the floating-point). Seems like there should be a isnani() function. I did play around with the f.nan == 1 (true) and f.nan == 0 (false) in testing, and it seemed to be correct at the time, but I guess I should double-check that logic again.I'm not sure what f.nan == true is supposed to test. Do you want to see if f is a nan? The isnan function in std.math is for that.I'm afraid it's a bad habit I've seemed to have picked up, "When in doubt...dup it!" Since Basic is my primary language and my D my secondary, I see the "out" in D as ByRef in Basic...and the fact that "s" is a local variable adds to the confusion (I have this great fear of a pointer referencing a memory location that disappears without warning...which has happen to me a few times in C).[snip]s2 )why the dup? You shouldn't need to change s so there is no need to make a copy. Also I can't tell exactly what string parse into complex numbers. For example will the string "1-2i" parse correctly? I think it should. In any case documenting what works and what doesn't is important.Well, Ben I'd like to thanks you for the time you spent looking things over! Did you have any time to test it in any of your D code? Thanks again. David L. ------------------------------------------------------------------- "Dare to reach for the Stars...Dare to Dream, Build, and Achieve!"
Apr 10 2005
"David L. Davis" <SpottedTiger yahoo.com> wrote in message news:d3d1bp$1gle$1 digitaldaemon.com...In article <d3cq53$19hh$1 digitaldaemon.com>, Ben Hinkle says...Your chances of getting something done to phobos goes up if you make it easy for him. It also makes it easier for us readers to have an idea what actually is intended to be in the final shipping library. I just assumed what you posted was what you expected to be accepted."David L. Davis" <SpottedTiger yahoo.com> wrote in message news:d3903a$vr8$1 digitaldaemon.com...Yeah, the comment style is what I started to use on code on my web-site (something I remembere picking up from my programming classes back in my college days), but I don't really expect Walter to keep them. :) And since I've passed all rights of the code to him, he can make the changes he needs to make.My first comment is the comment style. Does anything else in phobos use this comment style? Looking in std.conv I see comments like /************************* * Convert blah to foo. * Gammar: digit(digit) */ifind returns the int index of the first occurance of the requested char. No need to compare with a float.[snip]I thought it better to cast the numercial literal with a suffix ("f" for float). Is that the wrong way to do it? I could have used cast(float)0 instead, would this had been better?why 0f?Well, Ben I'd like to thanks you for the time you spent looking things over! Did you have any time to test it in any of your D code?no problem. I want to see phobos improve, too. Thanks for posting your change request. I didn't try running the code.
Apr 11 2005