Re: [tex-implementors] Ugly rules in PDF
I think ``horribly'' is a bit extreme. Rules are drawn like any other filled rectangle, and can be set to be anti-aliased in viewers which support that setting for graphic objects. Using a character is an interesting idea, but has difficulties as Hans noted. Just replacing it w/ a stroked line would be an alternative worth considering IMO. William
WillAdams@aol.com wrote:
I think ``horribly'' is a bit extreme.
Rules are drawn like any other filled rectangle, and can be set to be anti-aliased in viewers which support that setting for graphic objects.
Using a character is an interesting idea, but has difficulties as Hans noted.
get me right: replacing rules in math sounds ok to me, but not in other cases, which reminds me too much of slanted pictex lines built out of periods -) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
Hans Hagen
WillAdams@aol.com wrote:
I think ``horribly'' is a bit extreme. Rules are drawn like any other filled rectangle, and can be set to be anti-aliased in viewers which support that setting for graphic objects. Using a character is an interesting idea, but has difficulties as Hans noted.
get me right: replacing rules in math sounds ok to me, but not in other cases, which reminds me too much of slanted pictex lines built out of periods
True, math mode does not have the problem of selection, but using characters still sounds horrible. It *will* wind up being like pictex or that old sprite.sty in reverse. The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles. Add a few heuristics of height/width and you are set. -- Donald Arseneau asnd@triumf.ca
Donald Arseneau wrote:
True, math mode does not have the problem of selection, but using characters still sounds horrible. It *will* wind up being like pictex or that old sprite.sty in reverse.
The current state is quite a lot like latex's 'picture' environment. Of course drawing and converting the whole radical into a single glyph on-the-fly would be best, but that's a bit too much for the moment. Hartmut's idea of using a single scaled glyph is a dedicated internal font is quite promising.
The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles.
Not this one, apparently (pdftex already outputs stroked lines). Taco
On Wed, 18 May 2005, Taco Hoekwater wrote:
The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles.
Not this one, apparently (pdftex already outputs stroked lines).
i had temporarily disabled pdftex's pdf_set_rule heuristics (everything that has at least one dimension <= 1bp is stroked): There are no visual differences whether it's a stroked line or a filled rectangle. It seems that these rendering problems happen only later in the process. Regards, Hartmut
Hartmut Henkel wrote --
It seems that these rendering problems happen only later in the process.
That is a simple way of saying what I was trying to explain... You cannot deduce from the specification of a graphics object in a pdf file how it will be rendered; this will differ from device to device. chris
Taco Hoekwater wrote:
The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles.
Not this one, apparently (pdftex already outputs stroked lines).
fyi: there has been rendering problems with acrobat rules from the start, and pdftex has been adapted in the past to this: first rectangles, later lines or rectangles, depending on the size (there was/is a threshold); we did lot sof experiments (random rules, increasing steps, viewing, printing; i must have the test file somewhere on the attic) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
Taco Hoekwater wrote:
moment. Hartmut's idea of using a single scaled glyph is a dedicated internal font is quite promising.
a very cute idea indeed Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
WillAdams@aol.com wrote:
I think ``horribly'' is a bit extreme. Rules are drawn like any other filled rectangle, and can be set to be anti-aliased in viewers which support that setting for graphic objects. Using a character is an interesting idea, but has difficulties as Hans noted.
Donald Arseneau wrote --
The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles. Add a few heuristics of height/width and you are set.
Are we talking about rendering pdf here? Or rendering postcript? These are not the same (in general), especially when dealing with `rule-like objects'. To put it another way: pdf is far from device-independent:; this is an empirical fact that I have lived with for the last 5 years. chris
Chris Rowley wrote:
Are we talking about rendering pdf here? Or rendering postcript?
These are not the same (in general), especially when dealing with `rule-like objects'.
To put it another way: pdf is far from device-independent:; this is an empirical fact that I have lived with for the last 5 years.
it's just rendering pdf on the screen, for which adobe seems to implement different strategies each time (going back between old and new variants and such, either or not being influenced by anti-aliasing, cooltype, transparency correction, under/overprint, colorspaces etc); in print there are no problems (unless one uses a 96-144 dpi printer); Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
Hans
it's just rendering pdf on the screen,
Oh well, I never expect to see anything reasonable from maths on screen (and I spend a lot of time staring at just that right now as i am busy editing loads of it!). For us, there are far bigger problems with pdf for maths: no searching, no `screen readers', no `reflowing', ... But I agree that missing fraction bars can be a nightmare, especially in combinatorial number theory:-)
for which adobe seems to implement different strategies each time
and then there is everyone else's stategies!! chris
(unless one uses a 96-144 dpi printer);
So 145 dpi is OK?:-) (going back between old and new variants and
such, either or not being influenced by anti-aliasing, cooltype, transparency correction, under/overprint, colorspaces etc); in print there are no problems
Hans
----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
Chris Rowley wrote:
Oh well, I never expect to see anything reasonable from maths on screen (and I spend a lot of time staring at just that right now as i am busy editing loads of it!).
i hope fo ryou that in that case they gave you a 1920x1200 screen to work with since i'm not that sure if a ten-deep-nested root on a 80 char width page is that readable at all on the average screen
But I agree that missing fraction bars can be a nightmare, especially in combinatorial number theory:-)
in that case, i think that they are never missing, but just taking 15 pixels width instead of a few (ok, by making that rule transparent you can see through and still see the numbers that it covers -)
and then there is everyone else's stategies!!
lucky you that everyone is using tex for math, which is still better than staring at some presentation mathml rendered in a browser using funny symbols, scaling etc maybe, just as math visualization was adapted to paper / print, you need to adapt it for the screen /pdf; could be a nice talk of you for some tex conference: the New Math Visualization project (i bet that the EU will fund it esp when you mention that it serves the big ones like MS and Adobe) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
Donald Arseneau wrote:
Hans Hagen
writes: WillAdams@aol.com wrote:
I think ``horribly'' is a bit extreme. Rules are drawn like any other filled rectangle, and can be set to be anti-aliased in viewers which support that setting for graphic objects. Using a character is an interesting idea, but has difficulties as Hans noted.
get me right: replacing rules in math sounds ok to me, but not in other cases, which reminds me too much of slanted pictex lines built out of periods
True, math mode does not have the problem of selection, but using characters still sounds horrible. It *will* wind up being like pictex or that old sprite.sty in reverse.
The leap to characters is particularly excessive because most rule rendering problems can be relieved by using stroked lines instead of filled rectangles. Add a few heuristics of height/width and you are set.
nevertheless hartmuts suggestion to use a special built in rule char (type 1) and scale that horizontally or vertically makes sense since it then is rendered using the cooltype engine and proper hinting / antialiasing is done; in that case we have just one char for the horizontal bar in a root symbol (which itself is not much different than a bunch of strokes); it's a bit the best of both worlds i wonder how complicated it is to implement something like that so that we can see the results; i attached a sample file, the second line uses a scaled hyphen; i wonder how it renders different on low res screens (don't have one connected now) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
participants (6)
-
Chris Rowley
-
Donald Arseneau
-
Hans Hagen
-
Hartmut Henkel
-
Taco Hoekwater
-
WillAdams@aol.com