Re: Bug#881692: command-not-found: I re-wrote command-not-found
I re-wrote command-not-found in C. It consists of two C programs: command-not-found, which gets triggered by bash, and update-command-not-found, which digests the data obtained with apt-file update. AFAIK there is only one rough edge, in that the parsing of /etc/apt/sources.list is not the same as apt's parsing. I do not know enough C++ to use libapt to do this. https://github.com/shawnl/command-not-found/ -Shawn Landden -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
On Mon, Nov 13, 2017 at 11:50 PM, Julian Andres Klode wrote: > (forwarding this to ubuntu-devel-discuss and Zygmunt) > > On Mon, Nov 13, 2017 at 10:33:39PM -0800, Shawn Landden wrote: >> Package: command-not-found >> Severity: wishlist >> >> I re-wrote command-not-found to get rid of the python dependancy, and >> to reduce the database size, as to reduce memory usage. >> >> https://github.com/shawnl/command-not-found >> >> I was preparing to upload it to mentors as command-not-found-ng > > I also rewrote it years ago, but using the same database format, > just in C. It was a lot faster. I don't understand the memory usage > bit - it should not matter how large the database is, it's memory > mapped, and not read into memory, as such memory usage should be > roughly constant. > > Questions/Comments for your approach: > > * Did you test your format on a slow HDD with caches dropped? It > must not be slower than the Python one (that one is way too slow > already) - I did, it seems to be faster (0.4 vs 0.68 seconds) > - I believe the database-based C rewrite was even much faster, > though. I switched it to mmap() and am now getting 0.27-0.45 with caches dropped, even after adding translations. It is 100% C and sh. (same postinst and postrm) Ping. -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
> > > Ruby is just a major no go. Re-written in C. And in the future, what about Lua? It is only 300KB. > At that system level, the best choices > are Perl, Shell, and C++. Maybe Python (on Ubuntu it's in ubuntu-minimal, > but in Debian it's only used by standard priority and less, perl on the > other hand is required and essential). Ruby has the lowest priority > - optional. > > -- > Debian Developer - deb.li/jak | jak-linux.org - free software dev > Ubuntu Core Developer de, en speaker > -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
On Thu, Nov 16, 2017 at 11:39 PM, Xen wrote: > Julian Andres Klode schreef op 14-11-2017 8:50: > > * You should not depend on grep, sed, coreutils, they are Essential. >> > > Can I ask what this means? > > I actually assume that these dependencies are not *required*, not that you > can't use the tools. Required: yes. The highest priority. sysvinit was Required: yes until systemd came along https://www.debian.org/doc/debian-policy/#priorities Speaking of, I can't use 'apt-get indextargets' from shell and had to rewrite in ruby, because sed doesn't not support lazy matching, and I don't know how else to match NOT \n\n. (it also doesn't seem to support multiples of submatches.) Old regular expression implementations are showing their age (not to mention perl's non-regular features). -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
On Thu, Nov 16, 2017 at 6:44 PM, Colin Watson wrote: > On Thu, Nov 16, 2017 at 05:10:19PM -0800, Shawn Landden wrote: > > On Mon, Nov 13, 2017 at 11:50 PM, Julian Andres Klode > > wrote: > > > * It needs to be translated - also very important. > > > > I made a pot file and used translations from the python version, but I > > can't get my app to look for translations (as examined through strace). I > > read the gettext manual and do not know what I am doing wrong. > > Looking at > https://github.com/shawnl/command-not-found/blob/master/ > command-not-found.c, > your problem appears to be that you aren't calling setlocale(). You > should normally call this before calling bindtextdomain() and > textdomain(): > > setlocale(LC_ALL, ""); > > (The gettext manual does cover this, but possibly you were looking at > some different bit of it.) > Managed to re-use all the translations from launchpad of the existing command-not-found. > > -- > Colin Watson [cjwat...@ubuntu.com] > -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
> * Did you test your format on a slow HDD with caches dropped? It >> must not be slower than the Python one (that one is way too slow >> already) - I did, it seems to be faster (0.4 vs 0.68 seconds) >> - I believe the database-based C rewrite was even much faster, >> though >> > I tested with kyotocabinet backend and it was slower with dropped caches on a hard drive (1 second), which is the slow case I am most concerned with. Small makes a difference. The code is at https://github.com/shawnl/command-not-found/tree/kyotocabinet -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
On Mon, Nov 13, 2017 at 11:50 PM, Julian Andres Klode wrote: > (forwarding this to ubuntu-devel-discuss and Zygmunt) > > On Mon, Nov 13, 2017 at 10:33:39PM -0800, Shawn Landden wrote: > > Package: command-not-found > > Severity: wishlist > > > > I re-wrote command-not-found to get rid of the python dependancy, and > > to reduce the database size, as to reduce memory usage. > > > > https://github.com/shawnl/command-not-found > > > > I was preparing to upload it to mentors as command-not-found-ng > > I also rewrote it years ago, but using the same database format, > just in C. It was a lot faster. I don't understand the memory usage > bit - it should not matter how large the database is, it's memory > mapped, and not read into memory, as such memory usage should be > roughly constant. > > Questions/Comments for your approach: > > * Did you test your format on a slow HDD with caches dropped? It > must not be slower than the Python one (that one is way too slow > already) - I did, it seems to be faster (0.4 vs 0.68 seconds) > - I believe the database-based C rewrite was even much faster, > though. > Yes, as the disk IO is all the time, I think its best to keep the file size small. Then it has more chance of staying in memory. > * update-command-not-found should use apt-get indextargets > fixed > * You don't store components, hence you cannot tell people to enable > component. That's a very important use case for Ubuntu, where > not all components are enabled by default, but the database is > shipped in the package. > > You could just append / to each package name I think, > and strip it away when displaying. > fixed > * You should use getopt_long() to parse command-line options, and > support -h, --help :) > fixed > * pts_lbsearch belongs into usr/lib/..., not usr/share/... > the seperate binary is gone > > * You don't implement a closest matches function: > > $ command-not-found thunderbrd > No command 'thunderbrd' found, did you mean: > Command 'thunderbird' from package 'thunderbird' (main) > thunderbrd: command not found > $ ./command-not-found thunderbrd > thunderbrd: command not found > >This one is really important. People do make typos or misremember >command names, so the tool needs to be able to deal with that > >Should be easy to implement though, although you might have to >search multiple times - once for each alternative. All you need is > > def similar_words(word): > """ return a set with spelling1 distance alternative spellings > > based on http://norvig.com/spell-correct.html"""; > alphabet = 'abcdefghijklmnopqrstuvwxyz-_' > s = [(word[:i], word[i:]) for i in range(len(word) + 1)] > deletes= [a + b[1:] for a, b in s if b] > transposes = [a + b[1] + b[0] + b[2:] for a, b in s if > len(b)>1] > replaces = [a + c + b[1:] for a, b in s for c in alphabet if > b] > inserts= [a + c + b for a, b in s for c in alphabet] > return set(deletes + transposes + replaces + inserts) > > And search for what that returns. And you don't need to search for > those > at all if you have a direct match. > > fixed, and I believe bit-for-bit identical > * It needs to be translated - also very important. > I made a pot file and used translations from the python version, but I can't get my app to look for translations (as examined through strace). I read the gettext manual and do not know what I am doing wrong. > > * You need to Conflict with command-not-found and not Break AFAIUI > > fixed > * You should not depend on grep, sed, coreutils, they are Essential. > > fixed, now it uses ruby as my shell was hacky. > * You do have to Depend on apt-file, as that configures apt to download > the Contents files > > fixed > * You should not have identifiers starting with _ in the program, these > are reserved for the C implementation (like _cleanup_free_). > > fixed > Yes, and these are basically the same reasons my C prototype is > not in the archive. Also, I did not put a lot of work into it, as > I was waiting for PackageKit to take that over, but that was not > done yet. > > I think it's a worthwhile approach, and I can see it replacing > command-not-found if those tiny issues have been fixed. Then you > could also avoid the -ng moniker, and just take over the main > package (if Zygmunt does not mind), which also avoids a month > long NEW process :) > > -- > Debian Developer - deb.li/jak | jak-linux.org - free software dev > Ubuntu Core Developer de, en speaker > -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
Re: Bug#881692: command-not-found: I re-wrote command-not-found
On Nov 14, 2017 8:15 AM, "Julian Andres Klode" wrote: On Tue, Nov 14, 2017 at 03:35:02PM +, John Lenton wrote: > On 14 November 2017 at 12:34, Julian Andres Klode wrote: > > On Tue, Nov 14, 2017 at 01:00:54PM +0100, Zygmunt Krynicki wrote: > >> I would love if we have a compact representation of mapping from name > >> to list of bits of information where each bit can be a small structure > >> with some data. Apart from components for ubuntu archive it could be > >> used to store facts about snap packages, flatpaks, etc. I would try to > >> avoid a simplistic command -> package mapping as that will force us to > >> encode things into strings in an ad-hoc way. > > > > That makes sense to me. But then we're back on a db, I guess. I sort > > like this minimal approach. > > I was thinking in the other direction, was going to see how it behaved > with sqlite as the store. Would that be objectionable? Using a relational database for a simple key -> structure mapping seems overkill and a mismatch for the problem, and the SQL does not make it more readable. I'd play with lmdb and kyotocabinet, these are two high-performance key-value file databases and then encode a structure as mentioned before. I had some kyotocabinet code, (i maintain that package, which btw is in mentors) but this way is at least half the size. (Kyotocabinet is 1mb and it almost doubles the size of the db, even using lower overhead b-tree back end. These entries are just very small. For the text file approach, we can even go human, readable, like git: git just encodes a number in a fixed-length decimal number, we can do the same, and then just encode (length, key), (length, data) pairs after each other (or as mentioned, just use the "index" as the field id, and store field ids in the progrma). Uses a bit more space, but encodes everything in a format you could read with a text editor, and should not be terribly less efficient. The thing is: This needs to be as efficient as possible: it should be below 100ms (or better 50ms), regardless of whether caches are dropped or not. Python code | Shawn's code SSD, cache 50ms5ms SSD, " dropped 256ms 15ms HDD, cache 50ms5ms HDD, " dropped 530ms 15ms I guess Shawn's code could even be improved in performance by avoiding the subprocess execution, avoiding various ld cache lookups and library loads. I am going to have to bring it in process to add the spell check code. That said, space requirements might matter too. -- Debian Developer - deb.li/jak | jak-linux.org - free software dev Ubuntu Core Developer de, en speaker -- Ubuntu-devel-discuss mailing list Ubuntu-devel-discuss@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
