two suggestions to \typesetbuffer
Dear Hans, I have presentations that include buffers typeset with \typesetbuffer. One of the presentation includes 33 buffers, which are small ConTeXt samples. Any time I change anything on the presentation (not on the buffers themselves), every buffer is typeset again. And the compilation time is longer than probably desired. My first suggestion would be to skip compiling buffers again, if the following three requirements are met: 1. Buffer content (.tmp file, I guess) is identical (checked with a hash) with the previous one from last compilation. 2. PDF file hasn’t been changed since previous complete compilation (md5 sum again). 3. Buffer content was correctly compiled in previous compilation. My second suggestion may be a bit complex to implement, but I think it would be useful. Wouldn’t it be possible that the fonts are only embedded one in the presentation? Many thanks for your help, Pablo -- http://www.ousia.tk
On 11/8/2013 7:11 PM, Pablo Rodriguez wrote:
Dear Hans,
I have presentations that include buffers typeset with \typesetbuffer.
One of the presentation includes 33 buffers, which are small ConTeXt samples.
Any time I change anything on the presentation (not on the buffers themselves), every buffer is typeset again. And the compilation time is longer than probably desired.
My first suggestion would be to skip compiling buffers again, if the following three requirements are met:
1. Buffer content (.tmp file, I guess) is identical (checked with a hash) with the previous one from last compilation.
no temp file is used but a although a hash is possible it also means a lot of extra housekeeping due to the fact that buffers are reused (in which case no run happens)
2. PDF file hasn’t been changed since previous complete compilation (md5 sum again).
takes time so less to gain there
3. Buffer content was correctly compiled in previous compilation.
define correctly
My second suggestion may be a bit complex to implement, but I think it would be useful. Wouldn’t it be possible that the fonts are only embedded one in the presentation?
how often does it happen ... if we're talking about one page files processing should normally fast e.g. \setupbodyfont[pagella] \starttext \input tufte \stoptext takes .3 seconds on my laptop. normally fonts are embedded efficiently (subsets) and shared Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
On 11/08/2013 11:39 PM, Hans Hagen wrote:
On 11/8/2013 7:11 PM, Pablo Rodriguez wrote:
Dear Hans,
I have presentations that include buffers typeset with \typesetbuffer.
One of the presentation includes 33 buffers, which are small ConTeXt samples.
Any time I change anything on the presentation (not on the buffers themselves), every buffer is typeset again. And the compilation time is longer than probably desired.
My first suggestion would be to skip compiling buffers again, if the following three requirements are met:
1. Buffer content (.tmp file, I guess) is identical (checked with a hash) with the previous one from last compilation.
no temp file is used but a although a hash is possible it also means a lot of extra housekeeping due to the fact that buffers are reused (in which case no run happens)
Compiling the presentation I‘m talking about for the first time on my laptop takes about 210 seconds. Every subsequent compilation takes about 70 seconds. I know my laptop is old and slow. But if I make a copy of the presentation source and remove the \typesetbuffer commands, first compilation takes about 22 seconds and subsequent compilations take about 7 seconds. I have no idea on how ConTeXt works internally, but what I’m trying to say is that if \typesetbuffer has been already compiled, compiling again is no gain.
3. Buffer content was correctly compiled in previous compilation.
define correctly
I meant completely compiled: PDF has been generated.
My second suggestion may be a bit complex to implement, but I think it would be useful. Wouldn’t it be possible that the fonts are only embedded one in the presentation?
how often does it happen ... if we're talking about one page files processing should normally fast e.g.
\setupbodyfont[pagella] \starttext \input tufte \stoptext
takes .3 seconds on my laptop.
normally fonts are embedded efficiently (subsets) and shared
Sorry, there was a typo in my message: “that fonts are only embedded once in the presentation”. My second suggestion isn’t about compilation time, but about duplicated fonts. I tried to provide a sample, but I don’t know why it doesn’t work: \starttext These are bufers: \dorecurse{10}{\startbuffer[text:\recurselevel]\setuppapersize[A7]\starttext\input tufte\stoptext\stopbuffer\typesetbuffer[text:\recurselevel][frame=on]} \stoptext Of course, fonts are embedded fine in documents, but \typesetbuffer is a special case. Having the fonts embedded only once in the final document would be better. Many thanks for your help, Pablo -- http://www.ousia.tk
Am 09.11.2013 um 12:49 schrieb Pablo Rodriguez
On 11/08/2013 11:39 PM, Hans Hagen wrote:
On 11/8/2013 7:11 PM, Pablo Rodriguez wrote:
Dear Hans,
I have presentations that include buffers typeset with \typesetbuffer.
One of the presentation includes 33 buffers, which are small ConTeXt samples.
Any time I change anything on the presentation (not on the buffers themselves), every buffer is typeset again. And the compilation time is longer than probably desired.
My first suggestion would be to skip compiling buffers again, if the following three requirements are met:
1. Buffer content (.tmp file, I guess) is identical (checked with a hash) with the previous one from last compilation.
no temp file is used but a although a hash is possible it also means a lot of extra housekeeping due to the fact that buffers are reused (in which case no run happens)
Compiling the presentation I‘m talking about for the first time on my laptop takes about 210 seconds. Every subsequent compilation takes about 70 seconds.
I know my laptop is old and slow.
But if I make a copy of the presentation source and remove the \typesetbuffer commands, first compilation takes about 22 seconds and subsequent compilations take about 7 seconds.
I have no idea on how ConTeXt works internally, but what I’m trying to say is that if \typesetbuffer has been already compiled, compiling again is no gain.
Try the filter module: https://github.com/adityam/filter Wolfgang
On 11/9/2013 12:49 PM, Pablo Rodriguez wrote:
Of course, fonts are embedded fine in documents, but \typesetbuffer is a special case. Having the fonts embedded only once in the final document would be better.
You can try the beta ... i'm not sure about it yes but this one is faster, we might end up with several methods to choose from depending on usage. Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
On 11/10/2013 12:30 PM, Hans Hagen wrote:
You can try the beta ... i'm not sure about it yes but this one is faster, we might end up with several methods to choose from depending on usage.
Many thanks for the improvement, Hans. Compilation for the first time decreased from 210 seconds to 181 seconds. And subsequent compilations decreased from 70 seconds to 7 seconds. The sample shows the issue with embedded fonts: \starttext These are bufers: \startbuffer[text:1] \setuppapersize[A7] \starttext\input tufte\stoptext \stopbuffer \typesetbuffer[text:1][frame=on] \startbuffer[text:2] \setuppapersize[A7] \starttext\input knuth\stoptext \stopbuffer \typesetbuffer[text:2][frame=on] \stoptext The same font is embedded three times in the final PDF. I wonder whether it would be possible that the font would be only embedded once in the final PDF. Many thanks for your excellent work, Pablo -- http://www.ousia.tk
On 11/10/2013 1:32 PM, Pablo Rodriguez wrote:
On 11/10/2013 12:30 PM, Hans Hagen wrote:
You can try the beta ... i'm not sure about it yes but this one is faster, we might end up with several methods to choose from depending on usage.
The same font is embedded three times in the final PDF. I wonder whether it would be possible that the font would be only embedded once in the final PDF.
unless something is wrong in the engine, the actual shapes of fonts should be shared but it looks like the subsets are embedded anyway ... something to look into (not a context issue afaiks) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com | www.pragma-pod.nl -----------------------------------------------------------------
participants (3)
-
Hans Hagen -
Pablo Rodriguez -
Wolfgang Schuster