# namespace separator ideas - Thread: https://externals.io/message/20586 - Original archive: https://news-web.php.net/php.internals/20586 - Started by Greg Beaver on 2005-11-26T20:52:35+00:00 - Messages: 53 - Page 1 of 3: messages 1 to 25 (25 per page, add `?page=N` to the URL) - Next page: https://externals.io/message/20586.md?page=2 - Reply by email to internals@lists.php.net - More for agents: search with https://externals.io/search.md?q={query} · JSON API at https://externals.io/api · documentation at https://externals.io/llms.txt --- ## #20586 — Greg Beaver — 2005-11-26T20:52:35+00:00 URL: https://externals.io/message/20586#20586 Hi all, I have only one caveat with the \ separator, which is that it is a little bit too similar to division with /, and can result in some confusing code like: The same issue exists with all colon based separators (that sounds bad when read the wrong way...) because of the ternary operator, and :: with static classes/methods. class name1 { const name2 = 1; } $a = new name1:::name2; $b = name1::name2; // do you see the difference? I get confused ?> What about using the T_OBJECT_OPERATOR? This is a parse error in existing versions, and also implies some separation. class name1 { const name2 = 1; } $a = new name1->name2; $b = name1::name2; ?> I also proposed on IRC using \\ as this is similar to netware driver separators: However, I know Andrei hated this :). I very much prefer the use of ->, as this has the same advantage as :: of "rhyming" with current syntax. Greg --- ## #20587 — Jessie Hernandez — 2005-11-26T21:06:28+00:00 URL: https://externals.io/message/20586#20587 · In reply to: #20586 I personally don't like any of these, but I just thought of this one: "%%". Don't think it'll cause any problems at all, and look at the code: What do you think? ":::" is more intuitive for me, but "%%" is an acceptable alternative... Regards, Jessie Greg Beaver wrote: > Hi all, > > I have only one caveat with the \ separator, which is that it is a > little bit too similar to division with /, and can result in some > confusing code like: > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > > $a = new name1\name2; > $b = name1/name2; > ?> > > The same issue exists with all colon based separators (that sounds bad > when read the wrong way...) because of the ternary operator, and :: with > static classes/methods. > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > // this may be a parse error with the current namespace patch, > // but need not be if we use -> > class name1 > { > const name2 = 1; > } > > $a = new name1:::name2; > $b = name1::name2; // do you see the difference? I get confused > ?> > > What about using the T_OBJECT_OPERATOR? This is a parse error in > existing versions, and also implies some separation. > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > // this may be a parse error with the current namespace patch, > // but need not be if we use -> > class name1 > { > const name2 = 1; > } > > $a = new name1->name2; > $b = name1::name2; > ?> > > I also proposed on IRC using \\ as this is similar to netware driver > separators: > > define('name1', 1); > define('name2', 2); > > $a = new name1\\name2; > $b = name1/name2; > ?> > > However, I know Andrei hated this :). I very much prefer the use of ->, > as this has the same advantage as :: of "rhyming" with current syntax. > > Greg --- ## #20589 — Matt Friedman — 2005-11-26T21:19:18+00:00 URL: https://externals.io/message/20586#20589 · In reply to: #20587 Hi, I've been following your conversations all day and think I might have a nice idea for the namespace separator idea. Sorry to barge in on your conversation but I think the following has value. What about something like <- or <:: or even just <: It makes sense to me because a namespace is akin to a parent of the class and in a diagram I would use an arrow to point to the parent from the child. At least there is some type of parent <- child relation. examples: Namespace <:: ClassName or Namespace <- ClassName or Namespace <: ClassName Unless I'm missing something these symbols should not conflict with other ones. They also have more a PHP "feel". This is the most semantically pleasing for me and it mirrors a common convention and represents the actual relationship. In the language it is nice when the symbols have an appearance representative of the actual relationship. When I see \ I think of ugly Windoze file paths. Also ::: seems awkward somehow. Matt. > I personally don't like any of these, but I just thought of this one: > "%%". Don't think it'll cause any problems at all, and look at the code: > > $a = new name1%%name2; > $b = name1::name2; // I see the difference clearly > ?> > > > What do you think? ":::" is more intuitive for me, but "%%" is an > acceptable alternative... > > > Regards, > > Jessie > > > Greg Beaver wrote: > > […] > > -- > > > > -- -- Matt Friedman --- ## #20591 — Jared Williams — 2005-11-26T21:36:38+00:00 URL: https://externals.io/message/20586#20591 · In reply to: #20586 > Hi all, > > I have only one caveat with the \ separator, which is that it > is a little bit too similar to division with /, and can > result in some confusing code like: > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > > $a = new name1\name2; > $b = name1/name2; > ?> > > The same issue exists with all colon based separators (that > sounds bad when read the wrong way...) because of the ternary > operator, and :: with static classes/methods. > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > // this may be a parse error with the current namespace > patch, // but need not be if we use -> class name1 { > const name2 = 1; > } > > $a = new name1:::name2; > $b = name1::name2; // do you see the difference? I get confused ?> > > What about using the T_OBJECT_OPERATOR? This is a parse > error in existing versions, and also implies some separation. > > namespace name1 { > class name2{} > } > define('name1', 1); > define('name2', 2); > // this may be a parse error with the current namespace > patch, // but need not be if we use -> class name1 { > const name2 = 1; > } > > $a = new name1->name2; > $b = name1::name2; > ?> > > I also proposed on IRC using \\ as this is similar to netware driver > separators: > > define('name1', 1); > define('name2', 2); > > $a = new name1\\name2; > $b = name1/name2; > ?> > > However, I know Andrei hated this :). I very much prefer the > use of ->, as this has the same advantage as :: of "rhyming" > with current syntax. Hi, I was thinking -> too.. Had ideas of namespaces being object instances in their own right.. namespace foo { const ACONST = '1'; class bar {} } echo $foo->bar; // some mangled internal class name (foo_bar_123 or something) to make it unique $a = new $foo->bar(); $ref = new ReflectionClass($foo->bar); Jared --- ## #20592 — Jessie Hernandez — 2005-11-26T21:39:27+00:00 URL: https://externals.io/message/20586#20592 · In reply to: #20589 Hi Mat, Matt Friedman wrote: > Hi, I've been following your conversations all day and think I might > have a nice idea for the namespace separator idea. Sorry to barge in > on your conversation but I think the following has value. > > What about something like <- or <:: or even just <: > > It makes sense to me because a namespace is akin to a parent of the > class and in a diagram I would use an arrow to point to the parent > from the child. At least there is some type of parent <- child > relation. > > examples: > Namespace <:: ClassName > or > Namespace <- ClassName > or > Namespace <: ClassName > > Unless I'm missing something these symbols should not conflict with > other ones. They also have more a PHP "feel". > Hmm, "<:" doesn't look bad to me. I'd like to somehow conduct a poll to see what most prefer (it would be great if we could add polls on php.net). So far, I have these possibilites: ::: :> <:: <- <: > This is the most semantically pleasing for me and it mirrors a common > convention and represents the actual relationship. In the language it > is nice when the symbols have an appearance representative of the > actual relationship. > > When I see \ I think of ugly Windoze file paths. Also ::: seems awkward somehow. > > Matt. > Yeah, I feel the same way for "\". ":::" seems OK to me, but of course, I want to put what the majority prefer. Regards, Jessie --- ## #20593 — Jason Garber — 2005-11-26T21:39:44+00:00 URL: https://externals.io/message/20586#20593 · In reply to: #20586 Hello, I as a userland library author, would appreciate namespaces quite a bit. Here is an idea: namespace1..class() $x = new MyApp..MemberOrder(); -- Best regards, Jason mailto:jason@ionzoft.com Saturday, November 26, 2005, 3:52:35 PM, you wrote: GB> Hi all, GB> I have only one caveat with the \ separator, which is that it is a GB> little bit too similar to division with /, and can result in some GB> confusing code like: GB> namespace name1 { GB> class name2{} GB> } GB> define('name1', 1); GB> define('name2', 2); GB> $a = new name1\name2; GB> $b = name1/name2; ?>> GB> The same issue exists with all colon based separators (that sounds bad GB> when read the wrong way...) because of the ternary operator, and :: with GB> static classes/methods. GB> namespace name1 { GB> class name2{} GB> } GB> define('name1', 1); GB> define('name2', 2); GB> // this may be a parse error with the current namespace patch, GB> // but need not be if we use -> GB> class name1 GB> { GB> const name2 = 1; GB> } GB> $a = new name1:::name2; GB> $b = name1::name2; // do you see the difference? I get confused ?>> GB> What about using the T_OBJECT_OPERATOR? This is a parse error in GB> existing versions, and also implies some separation. GB> namespace name1 { GB> class name2{} GB> } GB> define('name1', 1); GB> define('name2', 2); GB> // this may be a parse error with the current namespace patch, GB> // but need not be if we use -> GB> class name1 GB> { GB> const name2 = 1; GB> } $a = new name1->>name2; GB> $b = name1::name2; ?>> GB> I also proposed on IRC using \\ as this is similar to netware driver GB> separators: GB> define('name1', 1); GB> define('name2', 2); GB> $a = new name1\\name2; GB> $b = name1/name2; ?>> GB> However, I know Andrei hated this :). I very much prefer the use of ->, GB> as this has the same advantage as :: of "rhyming" with current syntax. GB> Greg --- ## #20594 — dz@bitxtender.com — 2005-11-26T21:49:23+00:00 URL: https://externals.io/message/20586#20594 · In reply to: #20592 ::: is best. :> and <: look like my favorite smilies. No way I could ever use them in code ;) - David Am 26.11.2005 um 22:39 schrieb Jessie Hernandez: > Hi Mat, > > Matt Friedman wrote: > > […] > > Hmm, "<:" doesn't look bad to me. I'd like to somehow conduct a > poll to see what most prefer (it would be great if we could add > polls on php.net). So far, I have these possibilites: > > ::: > :> > <:: > <- > <: > > > […] > > Yeah, I feel the same way for "\". ":::" seems OK to me, but of > course, I want to put what the majority prefer. > > > Regards, > > Jessie --- ## #20595 — Stanislav Malyshev — 2005-11-26T22:06:17+00:00 URL: https://externals.io/message/20586#20595 · In reply to: #20589 MF>>Unless I'm missing something these symbols should not conflict with MF>>other ones. They also have more a PHP "feel". I wonder what "PHP feel" is in bunch of special symbols with meaning entirely obscure to non-initiated... Why not {@ or <* then? They are nice ASCII art too. And there are so many combinations of two special symbols, let's find a meaning for all of them. Or at least for as many as we can. -- Stanislav Malyshev, Zend Products Engineer stas@zend.com http://www.zend.com/ +972-3-6139665 ext.115 --- ## #20596 — Matthew C. Kavanagh — 2005-11-26T22:16:11+00:00 URL: https://externals.io/message/20586#20596 · In reply to: #20595 > I wonder what "PHP feel" is in bunch of special symbols with meaning > entirely obscure to non-initiated... Why not {@ or <* then? They are nice > ASCII art too. And there are so many combinations of two special symbols, > let's find a meaning for all of them. Or at least for as many as we can. Great, everyone loves sarcasm, but you're describing exactly the opposite of what's going on in this thread. There is a perceived need for this functionality and you could argue for or against the functionality instead. There's plenty of two character tokens in PHP already and people aren't calling it "PerlHP" yet. The -> suggestion if practicable seems the most intuitively correct to me. --- ## #20598 — Marcus Boerger — 2005-11-26T22:23:14+00:00 URL: https://externals.io/message/20586#20598 · In reply to: #20589 Hello Matt, bla! Saturday, November 26, 2005, 10:19:18 PM, you wrote: > Hi, I've been following your conversations all day and think I might > have a nice idea for the namespace separator idea. Sorry to barge in > on your conversation but I think the following has value. > What about something like <- or <:: or even just <: > It makes sense to me because a namespace is akin to a parent of the > class and in a diagram I would use an arrow to point to the parent > from the child. At least there is some type of parent <- child > relation. > examples: > Namespace <:: ClassName > or > Namespace <- ClassName > or > Namespace <: ClassName > Unless I'm missing something these symbols should not conflict with > other ones. They also have more a PHP "feel". > This is the most semantically pleasing for me and it mirrors a common > convention and represents the actual relationship. In the language it > is nice when the symbols have an appearance representative of the > actual relationship. > When I see \ I think of ugly Windoze file paths. Also ::: seems awkward somehow. > Matt. > > > […] > -- > -- Matt Friedman > -- > > Best regards, Marcus --- ## #20597 — dz@bitxtender.com — 2005-11-26T22:23:15+00:00 URL: https://externals.io/message/20586#20597 · In reply to: #20596 -> is crap because it is used for accessing object methods and properties. Come on, guys. It is not that difficult. Think before you write. - David Am 26.11.2005 um 23:16 schrieb Matthew C. Kavanagh: > > > […] > > Great, everyone loves sarcasm, but you're describing exactly the > opposite of what's going on in this thread. There is a perceived need > for this functionality and you could argue for or against the > functionality instead. There's plenty of two character tokens in PHP > already and people aren't calling it "PerlHP" yet. > > The -> suggestion if practicable seems the most intuitively correct to > me. --- ## #20600 — Jessie Hernandez — 2005-11-26T22:27:11+00:00 URL: https://externals.io/message/20586#20600 · In reply to: #20598 Marcus Boerger wrote: > Hello Matt, > > bla! > LOL! -- Jessie --- ## #20602 — Matthew C. Kavanagh — 2005-11-26T22:29:00+00:00 URL: https://externals.io/message/20586#20602 · In reply to: #20597 Don't let the identifiers collide, then. > -> is crap because it is used for accessing object methods and > properties. Come on, guys. It is not that difficult. Think before you > write. --- ## #20603 — Marcus Boerger — 2005-11-26T22:31:02+00:00 URL: https://externals.io/message/20586#20603 · In reply to: #20595 Hello Stanislav, i know that language it is call ed PerlHyPer :-) marcus Saturday, November 26, 2005, 11:06:17 PM, you wrote: MF>>>Unless I'm missing something these symbols should not conflict with MF>>>other ones. They also have more a PHP "feel". > I wonder what "PHP feel" is in bunch of special symbols with meaning > entirely obscure to non-initiated... Why not {@ or <* then? They are nice > ASCII art too. And there are so many combinations of two special symbols, > let's find a meaning for all of them. Or at least for as many as we can. > -- > Stanislav Malyshev, Zend Products Engineer > stas@zend.com http://www.zend.com/ +972-3-6139665 ext.115 Best regards, Marcus --- ## #20604 — dz@bitxtender.com — 2005-11-26T23:01:50+00:00 URL: https://externals.io/message/20586#20604 · In reply to: #20602 The proposal is just stupid. It cannot get any more confusing and inconsistent. Am 26.11.2005 um 23:29 schrieb Matthew C. Kavanagh: > Don't let the identifiers collide, then. > > > > […] --- ## #20606 — oliver.graetz@arcor.de — 2005-11-26T23:40:01+00:00 URL: https://externals.io/message/20586#20606 · In reply to: #20592 Jessie Hernandez schrieb: > So far, I have these possibilites: > > ::: > :> > <:: > <- > <: OK, now that Marcus threw a PostToLongException('bla'), I'd love to get back to this. Other suggestions were : :: -> .. \ Now some requirements: - needs to work without collisions in the parser (seems to kill ":","::","->") - should be easily distinguishable from other, similar operators (seems to kill ":::") - at least one person on earth should like it (seems to kill "\") This leaves us with: :> <:: <- <: .. I personally do not like ".." but that's just one voice. What about combinations with | (like ":|") ? For a laugh I should ass this link: The periodic table of Perl operators! http://www.ozonehouse.com/mark/blog/code/PeriodicTable.pdf OLLi --- ## #20610 — Matthew C. Kavanagh — 2005-11-27T00:09:40+00:00 URL: https://externals.io/message/20586#20610 > - looks OK (Pear%%Date) You're kidding, right? %% looks nothing but blocky and messy to my human eyes, and the resemblance to :: stretches the word "tenuous" to its limit. --- ## #20611 — Jasper Bryant-Greene — 2005-11-27T00:17:01+00:00 URL: https://externals.io/message/20586#20611 Oliver Grätz wrote: > Oliver Grätz schrieb: > > […] > > Whoops, forgot to add "%%" > +1 for <- here. It makes namespace hierarchies resemble diagrams of the same hierarchies (like inheritance in UML). It also visually makes sense, isn't hard to type, and doesn't look like a smiley (like :> etc). -- Jasper Bryant-Greene General Manager Album Limited +64 21 708 334 jasper@album.co.nz --- ## #20612 — Eric Coleman — 2005-11-27T00:28:47+00:00 URL: https://externals.io/message/20586#20612 · In reply to: #20611 I'd have to vote for either <- or :>. I perfer the later. \ is kinda crappy. Looks weird, as does %%. Looks like a mess. -- Eric Coleman > Oliver Grätz wrote: > > […] > > +1 for <- here. It makes namespace hierarchies resemble diagrams of > the same hierarchies (like inheritance in UML). It also visually > makes sense, isn't hard to type, and doesn't look like a smiley > (like :> etc). > > -- > Jasper Bryant-Greene > General Manager > Album Limited > > +64 21 708 334 > jasper@album.co.nz --- ## #20613 — dz@bitxtender.com — 2005-11-27T00:35:09+00:00 URL: https://externals.io/message/20586#20613 · In reply to: #20612 If we rule out :::, it should be <- instead. Foo:>bar:>Baz::getTest()->foo(); vs Foo<-bar<-Baz::getTest()->foo(); The latter looks better to me. I'd still prefer Foo:::bar:::Baz::getTest()->foo(); though. - David Am 27.11.2005 um 01:28 schrieb Eric Coleman: > I'd have to vote for either <- or :>. I perfer the later. > > \ is kinda crappy. Looks weird, as does %%. Looks like a mess. > > -- Eric Coleman > > > > > > […] --- ## #20614 — Robert Deaton — 2005-11-27T00:50:20+00:00 URL: https://externals.io/message/20586#20614 > If we rule out :::, it should be <- instead. I'd hate to see it be <-, looks too much like ->. Foo<-bar<-Baz::getTest()->foo(); but I agree that ::: should be ruled out all together, its too similar to ::. I'd almost rather have %% over <-, and even that looks kinda messy. I think if I had to, I'd pick :>, although there still isn't any that stands out to me as "Hey, this is definately a good choice" -- --Robert Deaton --- ## #20615 — Alan Knowles — 2005-11-27T01:09:09+00:00 URL: https://externals.io/message/20586#20615 · In reply to: #20614 This is getting kind of fun to read since we have unicode, perhaps we should use one of those characters ;) - I'm sure I can find a chinese character that means namespace for it ;) define('foo',1);define('bar',2);var_dump(foo<-bar); personally I would prefer ":" if it is at all feasible, (changing the parser to enforce whitespace on a ? b : c with a compile error would be a perfectly acceptable penalty. - It pales into insignficance compared to the `clone()` issue that has to be dealt with in upgrading PHP4->PHP5 ;) Regards Alan > > > […] > > I'd hate to see it be <-, looks too much like ->. > Foo<-bar<-Baz::getTest()->foo(); > > but I agree that ::: should be ruled out all together, its too similar to ::. > > I'd almost rather have %% over <-, and even that looks kinda messy. > > I think if I had to, I'd pick :>, although there still isn't any that > stands out to me as "Hey, this is definately a good choice" > > -- > --Robert Deaton --- ## #20617 — oliver.graetz@arcor.de — 2005-11-27T01:25:13+00:00 URL: https://externals.io/message/20586#20617 · In reply to: #20611 Jasper Bryant-Greene schrieb: > Oliver Grätz wrote: > > > […] > > > +1 for <- here. It makes namespace hierarchies resemble diagrams of the > same hierarchies (like inheritance in UML). It also visually makes > sense, isn't hard to type, and doesn't look like a smiley (like :> etc). If Marcus is correct, then this won't work because of such cases: define('a',2); define('a',3); if (a<-b) echo 'gotcha'; Sorry :-( OLLi --- ## #20620 — comex — 2005-11-27T02:02:12+00:00 URL: https://externals.io/message/20586#20620 (This is my first post to this list and of course just my opinion.) > Foo%%bar%%Baz::getTest()->foo(); > > Yes, it is blocky, as is the :: for class scopes. But it is the same width as *six* colons. I agree that : should be used even if it means requiring whitespace in some obscure case, mainly because everything else suggested here just doesn't look right. %%, as mentioned, is blocky, as with many of the other ones. And <- to me looks backwards and implies that x->y = y<-x. (Same with any of the operators starting with <, since that seems to be impossible anyway.) If : can't be used, however, %% is my favorite. --- ## #20625 — Eric Coleman — 2005-11-27T02:38:16+00:00 URL: https://externals.io/message/20586#20625 If : is still a viable solution, i'd much rather see that used :P -- Eric -- Eric Coleman > Great work, Oliver! I personally also add a +1 for ":", with the > whitespace restriction in the ternary operator (or just removing > namespace constants). > > I would also add another restriction: you cannot more than one > point for/against (so no +2's or -2's, etc.). It just doesn't make > sense (I might as well do +4000000 for whatever I want :-) ). > > > Regards, > > Jessie > > > > Oliver Grätz wrote: > > […] > > Namespace Operators > ====================================================================== > ================================================== > Name |s| %% | : | ::: | ::< | :< | <- | :> | > \ | -> | \\ | <:: | <: | .. | | | | > ---------------------+-+-----+-----+-----+-----+-----+-----+----- > +-----+-----+-----+-----+-----+-----+-----+-----+-----+ > Robert Deaton |-| +0 | | -1 | | | -2 | +1 > | | | | | | | | | | > Alan Knowles |-| | +1 | | | | | > | | | | | | | | | | > Greg Beaver |-| | +0 | +0 | +0 | +0 | | +0 | > +0 | +1 | +1 | +0 | +0 | | | | | > Jessie Hernandez |-| +1 | +1 | +1 | | | | > | | | | | | | | | | > Matt Friedman |-| | | -1 | | | +1 | | > -1 | | | +1 | +1 | | | | | > David Zülke |-| | | +2 | | | +1 | -1 > | | | | | -1 | | | | | > Oliver Grätz |*| +2 | | +1 | | | | > | | | | | | | | | | > Matthew C. Kavanagh |-| -1 | | | | | | > | | | | | | | | | | > Jasper Bryant-Greene |-| | | | | -1 | +1 | -1 > | | | | | -1 | | | | | > Eric Coleman |-| -1 | | | | | +1 | +1 | > -1 | | | | | | | | | > Christian Stocker |-| | | | | -1 | | -1 | > +1 | | | | -1 | | | | | > Stanislav Malyshev |-| | | | | | | > | | | | | | | | | | > Jared Williams |-| | | | | | | > | | +1 | | | | | | | | > Jason Garber | | | | | | | | > | | | | | | +1 | | | | > | | | | | | | | > | | | | | | | | | | > ---------------------+-+-----+-----+-----+-----+-----+-----+----- > +-----+-----+-----+-----+-----+-----+-----+-----+-----+ > sum | | +1 | +1 | +3 | +0 | -2 | +2 | -1 | > -1 | +2 | +1 | +1 | -2 | +1 | | | | > ====================================================================== > ================================================== --- Next page: https://externals.io/message/20586.md?page=2