I'm making a regex mistake. The regex works, but it's bad. I'm using grep. Here's the full command (trying to extract RPM filenames, basically it's basename-ing): grep -o -E "([[:alpha:]]*-)*([[:digit:]]\.*).*\.rpm"

What's a better way to extract filenames from paths with grep? Is grep a good choice for such thing?

@FaheemMitha I use it extensively. Among other interests, I run a publishing house, and SILE is our primary typesetting engine. It's really quite capable. I hack on it first to scratch personal itches and then a little more to push development forward as I am able. I think it has a lot of potential, but it also has some rough areas holding it back from mainstream adoption and I'm not sure when or how the handful of us working on it will overcome those.

By comparison I'm doing a spot of LaTeX work today and cursing it under my breath every minute. And that's not because I don't understand how to make it work for me –not to boast but I can make it work for it's keep– it's just that it's unpleasantly recalcitrant compared to SILE and knowing how things can be better it's hard to go back.

@Caleb I see. Interesting. I'm a TeX user, so I'm naturally interested in the competition. Which doesn't really exist, however. SILE is probably as close as anything. LOUT also in theory, but that doesn't seem like it is being developed.

Is the plan to keep actively developing it, then? And any plans to add math?

And how's performance in comparison to TeX? Which is supposedly pretty fast.

And is the use of Lua just a coincidence? As you're aware, there is also LuaTeX.

@FaheemMitha LOUT is dead as far as I know. Outside of the TeX family (XeLaTeX, ConTeXt, LuaLaTeX, etc, LaTeX 3, HarfTex, etc.) there really isn't much to turn too. Patoline is another interesting one, and to a some extent Platypus as well, but I'm pretty sure SILE is both already more versatile and holds a lot more potential than any other alternative.

@FaheemMitha Two people have separately gotten math working to some extent, we're working on merging their approaches now. There is a branch in the main repo that has a proof of concept. I don't know how production ready it is, but I think at least the basics of MathML at least render pretty well. I'm not sure how easy it is to integrate larger equations into other page layouts. Figure handling (floats, etc.) is one area LaTeX is still more powerful.

@FaheemMitha It's generally faster, but on the same general order of magnitude. If you run LuaJIT it can cut rendering times down to ½ or less.

A few things are slower, but the time taken is used for things LaTeX can't do such as optimizing future/past page breaks. Once LaTeX decides to eject a page it's done, and it's approach to memory reflects this. SILE keeps an eye of a few more things at once, and if you make use of that you can slow it down.

@Caleb What success have you had getting people interested in it? In theory there is a big market out there.

@FaheemMitha No it's not a coincidence. The typesetting related libraries floating around in Lua were an attractive point in that choice, although we don't use any LuaLaTeX stuff directly. Being a fast and easy to embedded language had more to do with that.

TeX works well, but it's a bit of a programming nightmare. And also frozen, so things aren't going to improve. And TeX 82 is around 40 years old, so showing its age in various ways. Not that I am an expert.

And conceptually it's pretty different that LuaLaTeX which still has TeX at its core and bolts Lua on top and exposes some of it's internals. Since SILE is Lua there is a lot more room to monkey with it on the fly than there ever will be in LuaTeX.

I wouldn't say it's never going to improve, the ecosystem just moves at glacial pace. There is some progress on LaTeX3 which modernizes a lot of things, but it's not ready for prime time either.

@Caleb Yes, I see. Though I tried LuaTeX, and didn't like the fact that the language is so minimal and DIY. Many things missing that I would consider basic, and half-baked stuff to fill in the gap.

@Caleb Meaning TeX? Actually, there has been a fair amount of activity in recent years, thanks to people like Joseph Wright, and Hans Hagen and the LuaTeX and ConTeXt teams.

But TeX is TeX and will always remain so.

@FaheemMitha SILE's support for non Latin scripts and layout issues is so much better we're seeing the most adoption interest from groups with projects in minority languages. People in the English/Latin/Cyrillic/Left-to-right-Top-to-botom worlds have generally figured out how to get by and are less eager to change tooling. But there are some that have too.

@Caleb I see. Is there any chance of this translating into development work? :-)

Oh, and have you thought about getting SILE into Debian? And/or other distributions/OS's? But Debian is not a bad place to start. That usually means more people at least trying it.

Some people are also more inclined to take software seriously if it is in Debian. I know I do.

Yes. Actually in the last couple months I've worked a lot of the build and release system so that packaging will be a lot easier. With the next release cycle I'm hoping a lot more distros get packages.

None of us developing are on Debian ourselves so it's a little hard to get that ball rolling, but asking for help packaging the dependency disaster it was a few moths ago would have been a rough place to start anyway.

@Caleb I see. I hope things continue to move forward. All too often projects falter and die.

BTW, I was a little puzzled that the creator of this project doesn't even mention it on his web site. If it's important to him, that's a weird way to promote it.

And I see you've opened a lot of the recent issues on GitHub.

Simon? He is a bit disorderly as far as his website(s) are concerned. They come and go and mention different areas of interest. Right now he's mostly into neural networks and automated font kenning, but you wouldn't know that from his website either.

@Caleb Yes, Simon. I don't know him, of course. I've just looked at his site.

So, does your publishing company use TeX too? I was looking at your activity on TeX SE. Complaining about LuaTeX being slow and so forth.

It's certainly slower than PDFTeX, that's for sure.

The book publishing is 100% SILE based. I have a few projects using XeTex, the final ones I haven't migrated to SILE are using either Lilypond or are just report systems spitting out lot of tables. Both of which have started to work in SILE but I haven't had the time to migrate something that was already working.

@Caleb Sounds busy. I suppose it's contract work?

A good way not to die is to have multiple developers. For example, Mercurial got interest quite early on, though initially it was just Matt Mackall's project. And it is still doing ok, though it's hard to foresee the future.

@FaheemMitha About the comments starred, sometimes you do have to compile things. In the past, either because of bugs, or because of old versions. Granted, it has been rarer nowadays. For instance, in my former job had to build debs of FreeRadius because it is badly supported in Debian.

Hi @RuiFRibeiro. How are you doing?

@RuiFRibeiro Yes, sometimes you have to compile things. But it's not common. Fortunately.

Like for instance, if you want some non-standard options.

Or if the maintainer messed up.

Or you need to patch something.

@FaheemMitha Oi. So so, thanks. Getting far thinner to see if some health issues improve. Not yet vegetarian again, but having a far more restrictive diet (again).

Yeah, FreeRadius used to have like a 9 year-old buggy package, and does not quite keep up with a fast moving target...

in Debian

@FaheemMitha Oh, and if installing assorted python modules outside of the distro packages, there is compiling involved. Unfortunately that happens a lot, especially when doing netops.

@RuiFRibeiro You could always package the modules. It's usually not hard. Particularly if you are planning to use it on multiple machines.

@MyWrathAcademia hi

@terdon hello

Thanks for moving to chat

So, what makes you think the single quotes aren't preserving the literal meaning?

When you send 'command1 && command2' to su -c, you are basically sending it a single command: 'command1 && command2'. It then launches a new shell session and executes that command. it is the shell that now sees the && and interprets them.

Essentially, you need the quotes to protect the command from your current shell. Without the quotes, su -c command1 && comamnd2 is interpreted as 2 separate commands: su -c command1 and command2.

Is that clearer?

So are you saying that 'command1 && command2' sends a string constant who the literal meaning of all its characters is preserved but then the new shell sees 'command1 && command2' and interprets && as a special character within that string?

@MyWrathAcademia Almost. It sends a string, yes. And then that string is passed to the shell which interprets it as a command.

Try it with set -x

$ su -c echo  a && echo b
+ su -c echo a

I get it now 'command1 && command2' preserves the literal meaning of all characters including && for the current shell. But the new shell (with a state) just receives command1 && command2. Is that a better way of understanding it?

@MyWrathAcademia Yep, exactly.

$ su -c 'echo  a && echo b'
+ su -c 'echo  a && echo b'

I meant (with a new state)

That's what you see above when you run it after running set -x. The first example only passes the first command to su because the && was intercepted by your current shell. While the second passes the entire thing to su because it was quoted.

I'm afraid I have to go have dinner now, but any of the other folks in this room will be able to explain at least as well as I can, most of them much better :)

Thanks for the excellent explanation

I have some more questions, hopefully someone can answer them

@MyWrathAcademia no; quotes are removed in the final stage of parsing; see gnu.org/savannah-checkouts/gnu/bash/manual/… for bash's (brief) description

@JeffSchaller thanks for answering. How long is the information in this chat room kept? Is it forever like on unix.stackexchange.com?

@MyWrathAcademia I think it's kept forever (the heat death of the universe, or Stack Exchange goes away, whichever comes first). You might like the transcript, which defaults to "today" but has links to go back in time, or the permalink for any individual chat message, available from the drop-down next to the message.

@JeffSchaller great to know. @terdon provided a link to this room. How do I get here from the main unix.stackexchange website and can you tell me where on this page the buttons to access the transcript and permalink are located?

@MyWrathAcademia ^^^

@JeffSchaller @JeffSchaller By the way, when you say quotes are removed in the final stages of processing, in the case of su - c 'sed -i "1i TEXT' file1.txt && sed -i "1i ANOTHER TEXT" file2.txt' do you mean only the single-quotes that enclose the single command intended for su -c? I ask because if all quotes are removed in the final stage of processing then the new shell created in the current shell would interpret the script (i.e. the argument) intended for sed not for the new shell

The argument that su receives is a single string sed -i "1i TEXT' file1.txt && sed -i "1i ANOTHER TEXT" file2.txt

@MichaelHomer thanks, just what I needed. Unix stack exchange just got even better. Can't believe I'm just finding out it's not just a question and answer site. So what kind of questions are allowed here? Just so I know what to discuss in chat

Questions should go on the site

@MichaelHomer, thanks for confirming.

@MichaelHomer Does su receive this single string argument in the new shell, even though su is called in the current shell?

@terdon I see @terdon already mentioned that su launches a new shell session and then executes that command (as root)

The su command receives it as-is and is not "in" any shell

@MichaelHomer I don't understand how su is not in any shell. Isn't the su command in the current shell since it is called in the current shell session? Please clarify

It received two string arguments that are exactly -c and sed -i "1i TEXT" file1.txt && sed -i "1i ANOTHER TEXT" file2.txt

It may then decide to launch a new shell as the root user and pass those along, or not

Is it possible to "cycle" a partition? E.g. I have a read-only VM with some writable areas, the writes only exist in memory, so when the VM is cycled everything goes bye-bye. Is there a way to do that to a partition? Say I have a partition and I want to use it to test a potentially malicious file, and want to be able to zero out that partition on demand.

Is there a way to reliably quarantine a partition and then dump its contents into the ether on demand? I'm guessing a chroot jail would play into that a little

chroot jail in that if the file is a malicious virus it may try to do something, but won't be able to reach outside of the jail

This suggests that the jail might not be effective as I think it could be: unix.stackexchange.com/questions/105/…

Side rant: MS Word sucks. I'm trying to get this paragraph to accompany the table it describes on the same page. If I have the table caption present, nothing works, if I delete the caption, the text does what it's supposed to do. I can't believe how horrid this stuff is. ugh