Live data from Hacker News

Show HN: FSQL – Search through your file system with SQL-esque queries

github.com

101–110 of 132 posts

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#101
post #25
post #11

Seems if the query is always going to start with SELECT, that maybe it should be assumed? I would never use this though, ack or find seem sufficient to me.

It's a shame Bash used 'select' as an elaborate menu built-in - it'd be quite neat to name the binary that (and drop the quotes). The you could just type the query right into your prompt!

"alias select=command select" lets you override it.

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#102

For those of us on Windows: Everything [1] does the job quite nicely with much less verbose syntax. [1] http://voidtools.com

Everything is pretty awesome, one of the things I really miss on Linux. When I last looked for a Linux replacement for it, none had the real time updates or the instant search ui, and even those that claimed to index the file system for a quick search were very slow. I actually ended up writing my own hacky and rudimentary GUI over locate to achieve something that fits my needs (and of course doesn't support real tim…

Try out KDE's search. It does realtime fulltext indexing and just works.

Also integrated with the KRunner framework.

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#103

For those of us on Windows: Everything [1] does the job quite nicely with much less verbose syntax. [1] http://voidtools.com

I've been using this for years and it is LIGHTNING fast. No need to "index" all the files because it reads directly from the MFT. If you create a new file matching the search pattern it's already sitting in the results window by the time you alt-tab. Also, the "directory size" equivalent of Everything is WizTree [1] ... much faster than WinDirStat, which I see recommended way too often. [1] http://antibody-software.c…

@all, do you guys know if osx / linux have the MFT like structures accessible to the user? Any links?

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#104

For those of us on Windows: Everything [1] does the job quite nicely with much less verbose syntax. [1] http://voidtools.com

I recently learned that you can also do advanced search (for e.g. date of modification) in Everything 1.4 (beta):

filename dm:>=16/05/2017 size:5mb..9mb parents:>=3

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#105

Earlier quoted context omitted.

That was my first thought, too. I'd love a file system that was more like a relational database. Also see the Pick operating system.

Transactional ACID updates to file systems would be pretty fantastic.

yes. thinking about this for a long time already

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#106
post #98

Does anyone remember WinFS (1)? Bill Gates described it as his biggest product regret (2). I remember I thought it was brilliant. Too bad it was probably a little bit too futuristic for its time, as for a few other things they launched when it just was not the right time... the clunky Tablet PCs (3) were for sure another example. (1) https://en.m.wikipedia.org/wiki/WinFS (2) http://www.zdnet.com/article/bill-gates-bi…

While WinFS was a good start, I think the idea of a semantic file system could be extended much further to the whole system (if it weren't for pesky POSIX) I think most people would expect /home/geokon/program1/src/ and /src/program1/home/geokon to have pretty much the same content A tag based file system that makes the two equivalent would eliminate all sorta of annoyances where you can't decide how to structure you…

There's a project that kinda wants to do that, ie provide a virtual interface on top of the existing filesystem based on tags you give each file: https://tmsu.org/

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#107

Earlier quoted context omitted.

Yeah, I like the idea of using sql but it is painful to write. It would be nice to omit the select and the quoting; my suggestion would be that fsql "select * from ..." could be written as select all from ... and all could be assumed if ommitted; so you could write select from or just from ... But unfortunately `select` is a sh(1) reserved word and `from` is an existing command! (shows who who your mail is from) So m…

> or just > from ... Or just ... Why even bother naming fields. Just make enter return everything from anywhere from all time for all people on all platforms. How many times are we going to have this ridiculous suggestion that less characters/words is automatically better. This is 'short/arrow functions' (pick your language) all over again, and invariably ends up with the situation where the new syntax is just fuckin…

still arrow functions are the single one ES6 addition to js that made the language finally nice enough for me to enjoy ;-)

as a fp geek

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#108

Earlier quoted context omitted.

I've been using this for years and it is LIGHTNING fast. No need to "index" all the files because it reads directly from the MFT. If you create a new file matching the search pattern it's already sitting in the results window by the time you alt-tab. Also, the "directory size" equivalent of Everything is WizTree [1] ... much faster than WinDirStat, which I see recommended way too often. [1] http://antibody-software.c…

@all, do you guys know if osx / linux have the MFT like structures accessible to the user? Any links?

Linux ext systems have inodes, which store all the metadata except the name. Directories contain name + inode number.

NTFS stores the name in the MFT instead, which makes hard links a bit weirder.

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#109
post #21
post #4

We have implemented smth. like this with sqlite extension, pretty powerful, with all the goodies sqlite (and its extensions) provides...

Sounds interesting, would love to take a look if it's available anywhere.

Unfortunately, it was commercial development, so it's not released. But implementation is relatively easy - it was a sqlite virtual table that (as much as I remember) looked in where condition for dir field, and listed that directory (= returned stat() data). Whole thing was quite interesting, because almost every component was somehow hooked into sqlite (either vith function or virtual table), so one could do pretty interesting things only with SQL.

Re: Show HN: FSQL – Search through your file system with SQL-esque queries

#110
post #98

Does anyone remember WinFS (1)? Bill Gates described it as his biggest product regret (2). I remember I thought it was brilliant. Too bad it was probably a little bit too futuristic for its time, as for a few other things they launched when it just was not the right time... the clunky Tablet PCs (3) were for sure another example. (1) https://en.m.wikipedia.org/wiki/WinFS (2) http://www.zdnet.com/article/bill-gates-bi…

While WinFS was a good start, I think the idea of a semantic file system could be extended much further to the whole system (if it weren't for pesky POSIX) I think most people would expect /home/geokon/program1/src/ and /src/program1/home/geokon to have pretty much the same content A tag based file system that makes the two equivalent would eliminate all sorta of annoyances where you can't decide how to structure you…

Years ago, I had a similar idea, but never did anything with it. Look at DNs in LDAP (and X.400/X.500), they are based on attribute=value pairs. What about a filesystem in which filenames were collections of attribute=value pairs?

e.g. /home/geokon/program1/src/foo.c

could become: user=geokon/program=program1/category=src/lang=c/name=foo

You could potentially decide that the order of the attributes is not significant, only the set of attribute-value pairs.

Downside: Too much typing. Although, maybe you could allow standard aliases for attribute names, so that:

user=geokon/program=program1/category=src/lang=c/name=foo

could also be written as:

u=geokon/p=program1/c=src/l=c/n=foo

(I think the attributes should be first-class filesystem objects, just like files are, as opposed to just text strings. Some of the values, e.g. an enumerated value like lang=c, should be first-class objects as well.)

I also took some inspiration (in concept not syntax) from https://en.wikipedia.org/wiki/Faceted_classification such as https://en.wikipedia.org/wiki/Colon_classification

Being too different from what everyone else is doing would be the real killer, however.

Post reply on HN