It does seem noticeably faster to me and makes the editing experience more pleasant.
Will TeXmacs get faster like this soon?
Will you switch from TeXmacs to Mogan due to its faster editing speed?
Mogan uses S7 and right now we have to patch it to make it work. My understanding of this patching is that it could introduce bugs. However seeing that Mogan does not seems to experience them we might consider to use S7 also in TeXmacs.
1 Like
I think so. S7 Scheme is a better choice than GNU Guile. And supporting two scheme implementation can free us from using only one scheme implementation. Since we can support two scheme implementation, I think it will not be a hard work to support the third one: Chez Scheme!
1 Like
In my previous benchmarks on the S7 TeXmacs branch, S7 was noticeably slower at converting a large TeXmacs file to LaTeX and was using about 30% more memory. Has this situation improved since?
1 Like
Did you find that editing large documents was noticeably faster using the S7 TeXmacs branch?
I’m not a user of LaTeX, for me I do not care about it am not a user of Export to LaTeX.
Other developers of Mogan should be interested in the LaTeX conversion, they will spare their time to improve the functionality and the performance!
For performance, as @amichail mentioned, editing is faster. I do think the performance of editing is much more important than conversion. For conversion, it is OK for us to wait a bit.
1 Like
I’m not really a user of LaTeX either, but the benchmark may be indicative of other performance issues unrelated to the LaTeX conversion. As a quick test I’ve just tried to load the Help -> Full manuals -> User manual in both TeXmacs and Mogan. Loading the document took roughly twice as long in Mogan. While inserting text at the end of the manual with Mogan, I got typesetting times of 2650ms, vs 975ms in TeXmacs.
2 Likes
How much memory does your computer have?
16Gb, that’s not the bottleneck, although Mogan does use about 30% more memory.
I think we need reliable benchmarks to understand the tradeoff of different interpreters. S7 as an interpreter is faster generally than Guile 1.8, but the use of scheme that TeXmacs does is very peculiar and there could be some problems in the way I implemented the tm-define overloading system which apparently does not play well with S7. I do not think there are bugs, but probably some optimisations are disabled due to the fact that we have to redefine some symbols.
2 Likes
Great! We will improve the performance later. Currently, I and jingkaimori are focusing on switching to the xmake build tool. Yufeng-shen just made the WASM port of mogan compile and work in browser (based on @mgubi’s work, see http://yufeng-shen.github.io/ ).
2 Likes
Do you think many programmers would be interested in the peculiar way TeXmacs uses scheme and the challenge this brings to changing scheme dialects? If so, maybe this issue could be used to bring attention to TeXmacs.
Who would pay for the servers used to provide the WASM port of Mogan?
If a link to such a port makes it to the top ten of Hacker News, would this result in a problem with cost and/or access to the port?
Currently, there is no server behind it. It is now on a Github static page.
Will people be able to run it directly from Github without paying anything for bandwidth costs?
In this case, Github pays the bandwidth costs of CDN.
Why does TeXmacs use a conservative subset of C++ but not a conservative subset of Scheme?
This does not work properly on my Firefox Android. The characters are rectangles.
Thanks, @darcy, I love your work on Mogan! I think it’s important we have a test bed for new developments that can then move to TeXmacs once they are stable.
What are your thoughts on strategies to make sure that Mogan and TeXmacs don’t diverge to a point where changes are incompatible?
3 Likes
Also, how should people refer to Mogan?
The problem with just saying “Mogan” is that, at this point, more people have heard of TeXmacs.
Also, Mogan is compatible with TeXmacs, so not mentioning TeXmacs is problematic in that respect as well.