> On 12/04/2012 11:45 PM, unruh wrote:
>> On 2012-04-13, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> On 12/04/2012 10:37 AM, John Wingate wrote:
>>>> Umm. You do know you only need one mv command and one specification of
>>>> the destination directory, don't you?
> If they're really named "file1", "file2", and "file3", maybe. But for
Uh, yes, duh. All I an saying is to amplify Wingates comments that one
can move a whole buch of files at once, not one at a time. As long as
there is some pattern to the files you want to move, you can glob your
way to moving them all at once.
> the general case you can't assume that the files that are to be moved > are exactly and only the ones matched by some wildcard pattern. What if > they're named "foo", "bar", and "baz" and there're a load of other files
mv foo bat baz path
> there that need to stay put? You might be able to cut it down to two mv > commands if "?a?" matches only the second and third of the files. Assume > there's a three-letter file to stay put, "qux", so "???" is no good.
> P.S. Please learn to trim. Your posts mostly end with a big block of > quoted text you aren't responding to, which might as well be deleted.
> Or does your primitive console-mode unix newsreader* make it a much > bigger hassle to do that than typing shift-ctrl-end, del in between > typing the period at the end of the last sentence of original text and > clicking "send"?
Nope. I just feel that if I am replying to a post, I will leave the
context for that reply there. What you consider extraneous the next
person might consider crucial. And I hate it when I get a reply so
trimmed I have no idea what the reply was to.
> * According to the headers of your post, you're using "slrn", so don't
> bother trying to deny the charge. :)
What charge?
And yes, I use a console-mode, unix newsreader. So what? Have you become
a member of thought gestapo now?
javax.swing.JSnarker <gharri...@boojum.mit.edu> wrote:
> unruh wrote:
>> javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> Robert Heller wrote:
>>>> They are. Have you tried them on a modern Linux system?
>> And what follows is a wonderful example of
>> [implied insult deleted].
Hey! Hey! Stop that!!! Why do you think I read these things???
> What does your classic unsubstantiated and erroneous claim have to do > with TeX, unruh?
-- John Forkosh ( mailto: j...@f.com where j=john and f=forkosh )
> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
>> On 12/04/2012 11:45 PM, unruh wrote:
>>> On 2012-04-13, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>>> mv file1 path/file1
>>>> mv file2 path/file2
>>>> mv file3 path/file3
>>> mv file? path
>>> will do them all at once.
>> If they're really named "file1", "file2", and "file3", maybe. But for
> Uh, yes, duh. All I an saying is to amplify Wingates comments that one
> can move a whole buch of files at once, not one at a time. As long as
> there is some pattern to the files you want to move, you can glob your
> way to moving them all at once.
And you can move them all at once in a GUI even when there isn't.
>> Or does your primitive console-mode unix newsreader* make it a much
>> bigger hassle to do that than typing shift-ctrl-end, del in between
>> typing the period at the end of the last sentence of original text and
>> clicking "send"?
> Nope. I just feel that if I am replying to a post, I will leave the
> context for that reply there. What you consider extraneous the next
> person might consider crucial.
If they want to reply to any of that they should reply directly to the post before yours.
> And I hate it when I get a reply so
> trimmed I have no idea what the reply was to.
Yes, that's erring too far in the *other* direction.
>> * According to the headers of your post, you're using "slrn", so don't
>> bother trying to deny the charge. :)
> What charge?
That you're using a primitive console-mode unix newsreader. Don't you know how footnotes work? :)
> And yes, I use a console-mode, unix newsreader. So what? Have you become
> a member of thought gestapo now?
What does your classic erroneous presupposition have to do with TeX, unruh?
On Thu, 12 Apr 2012 22:07:37 -0400, javax.swing.JSnarker wrote:
> On 12/04/2012 12:07 PM, unruh wrote:
>> On 2012-04-12, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> On 11/04/2012 11:27 PM, Robert Heller wrote:
>>>> They are. Have you tried them on a modern Linux system?
>> And what follows is a wonderful example of [implied insult deleted].
> What does your classic unsubstantiated and erroneous claim have to do
> with TeX, unruh?
Hmm. That sounds *very* familiar ...
-- The "reality TV" craze is as if your local hamburger joint replaced the burgers on their menus with tins of the cheapest noname dog food they could find in the pet store aisle of the supermarket, didn't change the prices, and magically continued to get just as many customers!
> On Apr 12, 10:20 pm, "javax.swing.JSnarker"<gharri...@boojum.mit.edu>
> wrote:
>>>> With a CLI, you need to specify the destination directory again and
>>>> again for each mv command you issue, instead of finding it just once.
>>> Umm. You do know you only need one mv command and one specification of
>>> the destination directory, don't you?
You do know, don't you, that there tend to be length limits on command line inputs? What if there are 30 files, or 300?
Also, that it's got to be typed into, in essence, a crummy line editor that makes it a pain to edit anything longer than one screen-width?
And no doubt if there's a typo in the name of the second file it will move the first, spit out an error message about the second, and leave the other 28 just sitting there, and you'll have to do nearly all of that typing again?
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> mv can move *multiple files* to a given directory in a *single* command.
Subject to command length limits and the cruddiness of trying to edit anything wider than the screen at that prompt ...
>> Er, by definition "what paste will insert" is "the contents of the
>> clipboard", so anything that changes the former clobbers the latter,
>> independently of whether there's a real, separate storage area used to
>> implement the clipboard or just a pointer into existing storage and a
>> length.
> Actually not necessarily under X11/Unix.
Actually yes. What part of "by definition" did you not understand?
And you seem to have the same issues with trimming unused quoted material as unruh. I take it either TkNews is as horrid for editing in as slrn is, or else it's some weird cultural thing in comp.text.tex?
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
On Friday, April 13, 2012 8:49:26 AM UTC+1, javax.swing.JSnarker wrote:
> On 13/04/2012 12:10 AM, Robert Heller wrote:
> > mv file1 file2 file3 path/
> > mv can move *multiple files* to a given directory in a *single* command.
> Subject to command length limits and the cruddiness of trying to edit > anything wider than the screen at that prompt ...
> >> Er, by definition "what paste will insert" is "the contents of the
> >> clipboard", so anything that changes the former clobbers the latter,
> >> independently of whether there's a real, separate storage area used to
> >> implement the clipboard or just a pointer into existing storage and a
> >> length.
> > Actually not necessarily under X11/Unix.
> Actually yes. What part of "by definition" did you not understand?
> And you seem to have the same issues with trimming unused quoted > material as unruh. I take it either TkNews is as horrid for editing in > as slrn is, or else it's some weird cultural thing in comp.text.tex?
> -- > public final class JSnarker
> extends JComponent
> A JSnarker is an NNTP-aware component that asynchronously provides > snarky output when the Ego.needsPuncturing() event is fired in cljp.
This thread has digressed so much that it's got nothing to do with comp.text.tex anymore. If you feel you still want to exchange opinion, may I suggest you move this thread to some comp.os group?
> This thread has digressed so much that it's got nothing to do
> with comp.text.tex anymore. If you feel you still want to
> exchange opinion, may I suggest you move this thread to some
> comp.os group?
You may *suggest* it ...
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> On 12/04/2012 11:56 PM, jon wrote:
> > On Apr 12, 10:20 pm, "javax.swing.JSnarker"<gharri...@boojum.mit.edu>
> > wrote:
> >>>> With a CLI, you need to specify the destination directory again and
> >>>> again for each mv command you issue, instead of finding it just once.
> >>> Umm. You do know you only need one mv command and one specification of
> >>> the destination directory, don't you?
> You do know, don't you, that there tend to be length limits on command > line inputs? What if there are 30 files, or 300?
The limit is fairly large. And as likely as not, there would be some
'pattern' to the names. Otherwise there is find(1), xargs(1), etc.
> Also, that it's got to be typed into, in essence, a crummy line editor > that makes it a pain to edit anything longer than one screen-width?
Actually not...
> And no doubt if there's a typo in the name of the second file it will > move the first, spit out an error message about the second, and leave > the other 28 just sitting there, and you'll have to do nearly all of > that typing again?
Actually not. Modern shells have command recall. Also with
auto-completion, I would not be typing all that much.
It is no worse with a GUI. One would still have to control-click on
each of the file icons and then one can still make mistakes doing that.
-- Robert Heller -- 978-544-6933 / hel...@deepsoft.com
Deepwoods Software -- http://www.deepsoft.com/ () ascii ribbon campaign -- against html e-mail
/\ www.asciiribbon.org -- against proprietary attachments
> On 12/04/2012 11:45 PM, unruh wrote:
>> On 2012-04-13, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> On 12/04/2012 10:37 AM, John Wingate wrote:
>>>> Umm. You do know you only need one mv command and one specification of
>>>> the destination directory, don't you?
> If they're really named "file1", "file2", and "file3", maybe. But for
> the general case you can't assume that the files that are to be moved
> are exactly and only the ones matched by some wildcard pattern. What if
> they're named "foo", "bar", and "baz"
What's wrong with mv foo bar baz path/
If the filenames are short, it's probably faster than trying to click
ctl-click ctl-click to accumulate the files, and then try to navigate
the directory sidebar to find the place to drag and drop them to,
without the pointer losing focus.
Actually, the biggest bugbear of computer systems for personal use is
their insistence on giving higher priority to their housekeeping tasks
than on watching the input device buffers. That means that half-way
through some complex edit, the system decides it needs to download
something, or rearrange its swap pointers, or reindex slocate, so it
grabs all the memory it can find, trashes the user's buffers, loses
focus on the mouse, fails to respond to any input, and heads off to do
its own stuff for a few seconds. I find this grossly unacceptable.
This is because the designers considered the system to be more important
than the user (understandably: it's their baby). But on a modern system
the user is the boss, and the clock should be reprogrammed to take that
into account. The system should never start or wake a background process
at a higher priority than user input.
One of the advantages of Unix-based systems is that you can fix that
behaviour (you can even nice a Mac, apparently). I have no idea what you
would do on Windows to stop it spending so much time reindexing the disk
every time want to do something, or spending more cycles doing a small
update than are consumed by an entire from-scratch installation of Linux.
> On 12/04/2012 11:56 PM, jon wrote:
>> On Apr 12, 10:20 pm, "javax.swing.JSnarker"<gharri...@boojum.mit.edu>
>> wrote:
>>>>> With a CLI, you need to specify the destination directory again and
>>>>> again for each mv command you issue, instead of finding it just once.
>>>> Umm. You do know you only need one mv command and one specification of
>>>> the destination directory, don't you?
> You do know, don't you, that there tend to be length limits on command
> line inputs? What if there are 30 files, or 300?
ARG_MAX is denominated in bytes (mine says 2097152) not numbers of
files. Anyway, what's wrong with using xargs?
If you need to do something like that with that many files then you
should find some common denominator and parameterise it in a script, or
treat it as a once-off task and do it by hand.
> Also, that it's got to be typed into, in essence, a crummy line editor
> that makes it a pain to edit anything longer than one screen-width?
Mine (Ubuntu) wraps perfectly happily line after line if needed, and
lets me use the arrow keys to edit it.
But yes, if you constantly find yourself worried by the terminal width,
you're either doin' it wrong or you need to seek professional advice.
> And no doubt if there's a typo in the name of the second file it will
> move the first, spit out an error message about the second, and leave
> the other 28 just sitting there, and you'll have to do nearly all of
> that typing again?
You really haven't used a modern Linux system, have you?
On 2012-04-13, Jail Bush Cronies <do....@ask.me> wrote:
> On 13/04/2012 12:19 AM, unruh wrote:
>> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
...
>>> * According to the headers of your post, you're using "slrn", so don't
>>> bother trying to deny the charge. :)
>> What charge?
> That you're using a primitive console-mode unix newsreader. Don't you > know how footnotes work? :)
I still do not see a charge. I see a statement, which, except for the
attempted pejorative "primative" is simply a statement of fact, not a
charge.
javax.swing.JSnarker <gharri...@boojum.mit.edu> wrote:
> On 12/04/2012 10:37 AM, John Wingate wrote:
>> javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> Particularly if you have a handful of them to move to the same
>>> destination directory. With a GUI you open a window at the source, open
>>> one at the target, click, ctrl+click, ctrl+click, ..., drag.
>> I find this to be slower and less convenient than a GUI+CLI combination.
>> YMMV.
> How so?
You don't have to open two windows, you don't have to drag, perhaps you
don't have to scroll the source window because the icons take up so much
real estate you can't see all of them at once, ...
>>> With a CLI, you need to specify the destination directory again and
>>> again for each mv command you issue, instead of finding it just once.
>> Umm. You do know you only need one mv command and one specification of
>> the destination directory, don't you?
>> In an xterm window you can list the files with ls, then construct the
>> argument list of files to be moved by selecting and pasting with the
>> mouse buttons. Fast (and saves typing long file names too).
BTW, in case you think unruh's response ("mv file? path") to your
example presents only two arguments to mv, it doesn't. The shell
invokes mv via "mv file1 file2 file3 path" (if those are the only three
files that fit the pattern).
> Sounds like that would be about the same as clicking and dragging files, > minus the ability to see the destination at the same time as the source. > (Or thumbnails, if picking and choosing picture files where that might > be a help.)
I don't see why seeing a destination window is an advantage. I grant
that thumbnails can be useful if the files have unhelpful names; e.g.,
PICT0001.JPG ... PICT0567.JPG.
>> Incidentally, selection in X11 doesn't necessarily copy anything
>> anywhere until it is pasted,
> Er, by definition "what paste will insert" is "the contents of the > clipboard",
Er, before you start defining things and presenting examples based on
your limited view, you should investigate the capabilities and
limitations of the X11 selection service. It's a different way of doing
things, with its own advantages and disadvantages. It has other uses than
manipulating blocks of text.
Incidentally, if you want to use clipboards in the way you are used to,
you can save selections to and retrieve them from any of several (eight,
maybe more) cut buffers.
> Copy text in open notepad window.
> Close notepad window.
> Click to position point in some other window.
> Paste => Crash! Tried to read from already-freed memory buffer.
Nope. Typically, no selection any more, so nothing to paste. If notepad
(X11 version :) ) and the target program behaved like xterm (and also used a
cut buffer) it would work as I think you would want.
Please stop criticizing what you don't seem to know very well on the basis
of what you think is the one true way of doing things.
-- John Wingate Mathematics is the art which teaches
joh...@tds.net one how not to make calculations.
--Oscar Chisini
javax> Now, if you ask me, the ideal UI would be an Explorer-like window with
javax> a little minibuffer added at the bottom that acts like a command
javax> prompt with that window's location in the filesystem as its current
javax> directory. (The cwd command there would, of course, make the whole
javax> window and not just the prompt go to the new directory.) Best of both
javax> worlds, then, as it would obviate the need to copy and paste the
javax> pathname in some instances, or to start a separate command interpreter
javax> window.
This - and a lot more - you could have on the Lisp Machine in the
Listener.
Those were the times.
'Andreas
-- ceterum censeo redmondinem esse delendam.
elzbet> On Thu, 12 Apr 2012 22:07:37 -0400, javax.swing.JSnarker wrote:
>> On 12/04/2012 12:07 PM, unruh wrote:
>>> On 2012-04-12, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>>> On 11/04/2012 11:27 PM, Robert Heller wrote:
>>>>> They are. Have you tried them on a modern Linux system?
>>> >>> And what follows is a wonderful example of [implied insult deleted].
>> >> What does your classic unsubstantiated and erroneous claim have to do
>> with TeX, unruh?
elzbet> Hmm. That sounds *very* familiar ...
Indeed, it does! And we are not even in comp.lang.lisp :-)
'Andreas
-- ceterum censeo redmondinem esse delendam.
> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
>> On 13/04/2012 12:19 AM, unruh wrote:
>>> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
> ...
>>>> * According to the headers of your post, you're using "slrn", so don't
>>>> bother trying to deny the charge. :)
>>> What charge?
>> That you're using a primitive console-mode unix newsreader. Don't you
>> know how footnotes work? :)
> I still do not see a charge. I see a statement, which, except for the
> attempted pejorative "primative" is simply a statement of fact, not a
> charge.
In other words, except for some little thing, it's not a charge. In other words, it *is* a charge. Gotcha.
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> On 13/04/2012 11:29 AM, unruh wrote:
>> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
>>> On 13/04/2012 12:19 AM, unruh wrote:
>>>> On 2012-04-13, Jail Bush Cronies<do....@ask.me> wrote:
>> ...
>>>>> * According to the headers of your post, you're using "slrn", so don't
>>>>> bother trying to deny the charge. :)
>>>> What charge?
>>> That you're using a primitive console-mode unix newsreader. Don't you
>>> know how footnotes work? :)
>> I still do not see a charge. I see a statement, which, except for the
>> attempted pejorative "primative" is simply a statement of fact, not a
>> charge.
> In other words, except for some little thing, it's not a charge. In > other words, it *is* a charge. Gotcha.
?? insults are not charges. They are just insults, especially when they
are lame insults. So, to parse it again. You made a statement of fact
into which you also threw in an attempted insult (primative). Except for
that attempted insult, the rest of your statement is simply a statement
of fact. With the insult it is a statement of fact plus and a lame insult. It
is still not a charge.
> If the filenames are short, it's probably faster than trying to click
> ctl-click ctl-click to accumulate the files,
I doubt it, unless they're one or two letter names, and there are possible length limits to the command line to consider. Also, you can typo a filename (and the longer the total length of the command line the proportionately more likely there'll be at least one typo) and the command will abort half-finished, whereas if you miss with a control click, you might move a file you wanted to keep and miss a file you wanted to move, but that's quick to correct and doesn't require redoing half the *rest* of the files. Just two.
Just don't let your finger slip off the ctrl key. :)
And then there's the case where you want to move, say, all the pictures of Tom's face. It's easy to pick those out with a thumbnail view in Windows Explorer or something similar, and just control-click them as you see them and then move them. Contrast Unix:
* ls
* my-previewer foo.jpg
* my-previewer foo2.jpg
* my-previewer bar.jpg
* Aha, Tom's face!
* mv bar.jpg dest/
* my-previewer quux.jpg
* my-previewer mumble.jpg
* mv mumble.jpg dest/
* my-previewer frotz.jpg
* Uh oh, next name after frotz.jpg scrolled off the top of the screen
* ls
* my-previewer quuuux.jpg
...
> Actually, the biggest bugbear of computer systems for personal use is
> their insistence on giving higher priority to their housekeeping tasks
> than on watching the input device buffers. That means that half-way
> through some complex edit, the system decides it needs to download
> something, or rearrange its swap pointers, or reindex slocate, so it
> grabs all the memory it can find, trashes the user's buffers, loses
> focus on the mouse, fails to respond to any input, and heads off to do
> its own stuff for a few seconds. I find this grossly unacceptable.
Most GUIs do seem to have issues in that area. They also don't seem to queue up keyboard and mouse events in a common buffer, so if the system gets slow, you can shift-click and it will act like you clicked and then tapped shift, or similarly.
This is not an inherent problem with the GUI concept, though; just a problem of shoddy implementation of the concept.
But there are ways to tame the beast -- at least on Windows.
> One of the advantages of Unix-based systems is that you can fix that
> behaviour (you can even nice a Mac, apparently).
Macs these days run a disguised version of BSD Unix so that's hardly surprising.
> I have no idea what you would do on Windows to stop it spending so
> much time reindexing the disk every time want to do something, or
> spending more cycles doing a small update than are consumed by an
> entire from-scratch installation of Linux.
Disable indexing service.
Disable other unnecessary services and startup items. (There are Web sites that can guide you regarding what's junk, what's needed only for XYZ so is junk if you don't need XYZ, and what's important and/or not prone to being a pain.)
Disable automatic updating so it just notifies you, and then install updates at *your* convenience after said notification.
Get Sysinternals ProcessExplorer (it's free, though not libre) and use it to lower the priority of anything that shows as gobbling excess CPU when you notice the interface get slow or jerky.
Unfortunately, there are some ill-behaved things that seem able to bypass the priority controls in Windows. Usually these aren't Microsoft programs but third-party ones. The file-sharing software Shareaza is a known instance -- it can occasionally slow the system down even when given the lowest priority, but this only tends to happen if you have a giant library of over 30k files or so and a new one downloads. Whatever it does is also quadratic in the size of the entire library, so it's obvious they used bubblesort somewhere and on a thread with high priority compared to the application. No easy way to fix thread priorities even with ProcessExplorer.
Another pain is that, while ProcessExplorer will tell you which of the fifty svchost.exes suddenly started eating a full core and will tell you via tooltip what services are in a svchost container, it won't tell you which service is the one responsible for the CPU spike. Sometimes there's only one; Windows Defender has a svchost.exe all to itself and randomly grabs a full core for a few minutes every few hours for no discernible reason. Other times it's a svchost with a dozen or more hosted services and it's anybody's guess which one's the culprit.
That said, it's not exactly all sweetness and light on the Unix side either. Beside the much more cumbersome interface (CLI "top", then "ps", then "nice", copying and pasting long opaque numeric processIDs all the way, vs. a neat list of processes you can keep sorted by CPU% and just right-click "priority Idle" the topmost item from) there's also the minor matter that, again, the system can start whole new processes with high priorities. You can nice them, but next time it comes back it will have the original high priority again. This can happen on both systems, of course, but starting whole new processes to do everything instead of having long-running mostly-idle processes is much more a Unix thing than a Windows thing.
Well, except for one area: installs involving .NET. If you install a .NET app or framework update, and watch in ProcessExplorer, you'll see a ton of short-lived processes starting up -- TrustedInstallers and a bunch of other things with more obscure names. If you poke and prod them you'll discover they're a compiler and related tools. In other words, when it installs something .NET does something rather reminiscent of tar -xf foo, make, make install as far as the system processes are concerned.
Fortunately, it only does this when you tell it to install something -- given that you have disabled fully-automatic installs, of course.
> At Fri, 13 Apr 2012 03:46:49 -0400 "javax.swing.JSnarker"<gharri...@boojum.mit.edu> wrote:
>> On 12/04/2012 11:56 PM, jon wrote:
>>> path/ now contains file1 file2 file3
>> You do know, don't you, that there tend to be length limits on command
>> line inputs? What if there are 30 files, or 300?
> The limit is fairly large. And as likely as not, there would be some
> 'pattern' to the names. Otherwise there is find(1), xargs(1), etc.
More annoying extra commands to learn just to work around the fact that something scales poorly. Vs. open destination folder alongside source folder and drag, drag, drag, which is faster than typing random long names and scales linearly in the total number of files (and NOT in the lengths of their names). Or, when there is a pattern, using search or sort order to get those files in a solid batch and then click, shift-click, drag.
>> Also, that it's got to be typed into, in essence, a crummy line editor
>> that makes it a pain to edit anything longer than one screen-width?
> Actually not...
Actually yes. Anything longer than 80 characters = horizontal scrolling. Without even the mitigation of a horizontal scrollbar being available. To go left and change something you have to hold down <- for ages; you can't just zip rapidly to the left and click at the site of your intended edit.
And you're not even guaranteed that basic editing functions will work. I've seen all kinds of misbehavior in these circumstances:
* Hitting left-arrow rapidly spams the command line with repetitions of
some bit of nonsense text like ^]]D^]]D^]]D.
* It works, but one or both of bksp, del does not DTRT.
* Typing overtypes rather than inserts and INS doesn't fix it, etc., so
if you need to replace a shorter item with a longer one you're
attached to something by an inclined plane wrapped helically around
an axis, so to speak.
* In the worst case the line is not editable at all; left arrow, bksp,
del, right arrow ALL fail. If you flub something, at best you can hit
enter and start over. And if you go to type "rm *.o" and suddenly
notice that you've accidentally gotten "rm *>o" instead, the only
way to save yourself is to *disconnect* or (if you're at the machine
in person) *hit the goddamn power switch*! Unbefuckinglievable!
Worse, this seems to be highly dependent on the particular Unix and even the particular installation of that particular Unix. Two machines both running the same version of Solaris that you sometimes have occasion to telnet to might differ, one supporting all the normal editing keys at its command prompt and the other not supporting left-arrow.
The one thing you *can* rely on in that area is that nothing in the way of editing keys will work correctly at the unix login prompt. ^]]D, etc. If you typo your username, you need to hit enter, hit enter, stare at the "Login incorrect" message for the several seconds' delay that's there to limit the rate at which anyone can attempt to brute-force an account's password, and then start over.
>> And no doubt if there's a typo in the name of the second file it will
>> move the first, spit out an error message about the second, and leave
>> the other 28 just sitting there, and you'll have to do nearly all of
>> that typing again?
> Actually not. Modern shells have command recall.
If the system you're using is lucky enough to have a modern shell. And working editing keys, without which command recall isn't very helpful. And even then you have roughly as many left-arrow keypresses to make as you would have had filename characters to retype anyway.
> It is no worse with a GUI. One would still have to control-click on
> each of the file icons and then one can still make mistakes doing that.
Unless you made a non-ctrl click by mistake, you can correct it easily by control-clicking the same item again, and then the intended item. You can also avoid building up a big, complex, delicate selection in any of three ways:
* Do the moves/etc. one by one. You'll need to anyway if you want to
give them all ad hoc new names at the same time, and it's not much
slower if you position the destination side by side with the source.
* Sort the files, if there's a characteristic that groups them in one
batch, and click shift-click to select a range. The intended files
may have a characteristic in common, such as all being 42 bytes in
size, all having names starting with "foob", all being modified
on June 17, 2002, or even all being 1024x768 jpegs, and Explorer
sort options can bunch them together.
* If there's a more complex naming pattern than a common prefix you
may be able to use search to generate a results window with exactly
the desired files, and then use that as the source of the drag
operation. Now you can even just hit ctrl-A to select all.
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> On 13/04/12 08:46, javax.swing.JSnarker wrote:
>> You do know, don't you, that there tend to be length limits on command
>> line inputs? What if there are 30 files, or 300?
> ARG_MAX is denominated in bytes (mine says 2097152) not numbers of
> files.
Not every system has it that large. Lots I've seen balk at a command line of over 256 characters -- easily exceeded by a *single* filename with a deeply-nested path.
> Anyway, what's wrong with using xargs?
More typing. Another command to learn. More complexity. Vs. simply control-clicking one more file.
Which would you prefer, a car like typical ones now or one where, when you go past 40mph, you suddenly need to use the wheel, pedals, and stick completely differently to do the same things as you did below 40?
> If you need to do something like that with that many files then you
> should find some common denominator and parameterise it in a script, or
> treat it as a once-off task and do it by hand.
And for "doing it by hand" a GUI is ideal.
>> Also, that it's got to be typed into, in essence, a crummy line editor
>> that makes it a pain to edit anything longer than one screen-width?
> Mine (Ubuntu) wraps perfectly happily line after line if needed, and
> lets me use the arrow keys to edit it.
Oh, that's even *worse*! It sets up the expectation that if you want to change something directly above the cursor you can just hit up-arrow and change it, and then you hit up-arrow and the whole thing disappears, replaced with the previous command, and then you hit down-arrow and get either a blank prompt or the command you had been editing -- but with your edits undone.
> But yes, if you constantly find yourself worried by the terminal width,
> you're either doin' it wrong
Yeah, you're doin' it wrong, alright, and the error you're making is called "not using a GUI"! That's a sure sign that your best option is to use a GUI to hands-on *grab* those files and *haul* them from place to place instead of trying to order the brain-dead butler called "the shell" to do it for you. Unless it's an oft-repeated task, where you might want to consider programming something to do it for you instead that has a simple way of executing it, like
% tar-up-new-image-files
instead of "ls -whatever-bloody-option-sorts-by-modification-date" <enter>, tab to other term window, "tar -xf ", and then lots of tabbing back and forth copying and pasting file names.
> or you need to seek professional advice.
Most mere mortals do NOT have a Unix guru on call 24/7, so that's simply not an option for any system you're trying to put forward as "user-friendly".
>> And no doubt if there's a typo in the name of the second file it will
>> move the first, spit out an error message about the second, and leave
>> the other 28 just sitting there, and you'll have to do nearly all of
>> that typing again?
> You really haven't used a modern Linux system, have you?
I already answered that question (in the negative) in an earlier post. That one had command history on up-arrow and the editing keys worked in the command line, but as I mentioned it had lots of other problems, ranging from documentation to lack of a driver for the 56K modem (no broadband in my area then) and thus no network access(!) to a lot of things not working out of the box (after reboot to Windows, search for rpms online, download, reboot to Linux, run command to mount NTFS volume, cd to download directory, run rpm...), to the way the system seemed to suffer bitrot on any unplanned shutdown-and-reboot (and this area gets electrical storms a lot in the summer, so there are a lot of 2-sec power outages with occasional longer ones).
The latter two were particular nuisances. An rpm for a given application would typically complain about missing dependencies. So I had to copy these, paste into editor, save into file on NTFS volume, reboot, pick Windows from boot menu, go to rpm archive sites searching for the dependencies in the list, download them, once they'd all downloaded (at 56K!) reboot, pick Linux from boot menu, try to install dependencies with rpm, get lists of *their* dependencies ... At the end of the process, there was maybe a 50/50 chance the thing worked OOTB. If it didn't, it often complained that files weren't in some place it was expecting them, but sometimes was less clear why it wasn't working. Files in wrong places could generally be fixed by finding them, moving them if they weren't used by anything else, or copying them or making symlinks if they were. I think there was another 50/50 chance or so there, so there was a 50% chance the application would work OOTB, a 50% chance it could be made to work with some extra manual steps post-install, and a 50% chance it was just plain broken, or at least had undocumented extra install steps that I'd have needed that 24/7 on-call Linux guru to discover and perform.
As for the power-failure bit-rot, the cause of *that* (but not the fix, short of lighting a fire under the dev-team's ass and then waiting years for an upgrade to ship) was obvious: OS misdesign, to wit, caching disk writes and not flushing the cache at all quickly even when the system was idle. No doubt born of Unix's heritage as something run on mission-critical servers and mainframes that would a) pretty much never be idle during production and b) all have big expensive battery backed up power supplies in their big expensive corporate or university machine rooms.
Someone ported Unix to the desktop without *thinking*. Gah.
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>> On 12/04/2012 10:37 AM, John Wingate wrote:
>>> javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>>> Particularly if you have a handful of them to move to the same
>>>> destination directory. With a GUI you open a window at the source, open
>>>> one at the target, click, ctrl+click, ctrl+click, ..., drag.
>>> I find this to be slower and less convenient than a GUI+CLI combination.
>>> YMMV.
>> How so?
> You don't have to open two windows, you don't have to drag, perhaps you
> don't have to scroll the source window because the icons take up so much
> real estate you can't see all of them at once, ...
If the file selection is ad-hoc, you need to have more information than just the name for each one, typically, when choosing them, AND the time to decide, for each file, whether to move it typically dominates the time taken to do one more control-click or even full click-drag operation.
When the computer's automation a) lets you do the job faster than you can think about each item and b) doing so involves a hairy syntax full of punctuation characters and case-sensitive text, massive and catastrophic errors that happen too fast for you to stop them become likely in the long run. :)
If the file selection isn't ad-hoc, the GUI, the CLI, or both usually offers some alternative method, such as wildcards or search results or sort order, for grabbing the desired files as a unit.
As for opening two windows:
a) It's not very difficult or slow.
b) It only takes about as much work as specifying each subdirectory of
its path at the CLI does.
c) You can see what you're doing, so if you go astray you realize it
immediately. GUI: double-click "foo", double-click "bar", see
"local" and "locals" folders, double-click "local", don't see
"baz", so back button, double-click "locals", and move on. CLI:
% mv file1 file2 file3 /foo/bar/local/baz/quux/and-so-on
no such path /foo/bar/local/baz
% *$&@!!!!
d) You can see the full context of your command, to wit, the contents
of both directories and not only the source (if even that). In
particular, before copying you can see if you already did it earlier
and forgot; you can see if you're in the wrong directory (from an
unexpected set of visible files) despite the path looking, at first
glance, correct. (Maybe both local and locals have the baz/quux/and-
so-on subpath in the example above.)
So, there are benefits, in the form of additional opportunities to spot and avoid or recover from errors, and no real costs, to that opening-another-window bit.
In particular, in the example above with "local" and "locals", you can detect the error instantly and fix it with two clicks and keep going. After
% mv file1 file2 file3 /foo/bar/local/baz/quux/and-so-on
no such path /foo/bar/local/baz
your best bet is up-arrow and then a whole lotta left-arrows to the end of "local", then type the "s". That's slower than two mouse clicks if there's very many left-arrows needed at all, AND it depends on command history being a feature of that shell, which in my experience is by no means guaranteed.
>> I was talking about moving multiple files.
> So was I. Your [insult deleted] is showing.
What does your classic unsubstantiated and erroneous claim have to do with TeX, Wingate?
59 input gestures. Some command prompts have a command length limit as low as 256 characters, so quadruple the number of files or give them longer names and you have a problem.
Versus control-clicking scaling and requiring just five input gestures: click aaa.jpg, control-click bbb.png, control-click, control-click, drag and drop on dest_dir which is presumably a subdirectory of the current dir.
If it's not, then getting at it takes another input gesture per subdirectory to expand from its deepest common superdirectory with the current one, versus many more at the CLI for all the path elements from root or else a whole bunch of ../../s.
Completion, *if* supported at that shell, *might* reduce that to triple or quadruple the work of control-clicking as you can then just type enough of a prefix to uniquely specify a file or directory and then hit tab for each of the GUI way's mouse clicks.
Copy and paste reduces it to six times the work of the GUI solution, to wit, for each GUI control click you have a GUI/CLI click, shift-click, copy, click, paste, space. If words select on double-click (de rigueur in any decent text editor; I wouldn't think it could be relied on in a terminal whose GUI is a bolted-on afterthought) then it's a mere five times as much work: double-click, copy, click, paste, space instead of control-click.
(The extra work involved in scrolling, if the source directory's contents won't fit on one screen, is one wheel flick for every N selected files, so it's lost in the noise. Besides, an ls output you're copying from and a "list" view in Explorer will be about equally compact, and the ls output *can't* be scrolled, you have to rerun ls with a pattern to exclude the first part of the alphabet.)
>> Sounds like that would be about the same as clicking and dragging files,
>> minus the ability to see the destination at the same time as the source.
>> (Or thumbnails, if picking and choosing picture files where that might
>> be a help.)
> I don't see why seeing a destination window is an advantage.
See above. Even seeing the stops along the way in navigating *to* the destination in a window can be an advantage.
> I grant that thumbnails can be useful if the files have unhelpful names;
> e.g., PICT0001.JPG ... PICT0567.JPG.
Guess how files tend to be named by digital cameras and/or the tools that import them via USB onto your computer's disk.
Files acquired from P2P or Usenet are prone to long and/or awkward and/or just plain incorrect naming.
Files that are normally created and dealt with by automation, ditto. Node301845.dat, anyone? And if you don't think the need occasionally arises to fix something or back something up manually with such files, dream on. Sure, those won't be helped by thumbnails, but they *sure* won't be helped by shell wildcards either. Control-click or sort-by-modified-date or similar it is, then, and the GUI shines in this case.
>>> Incidentally, selection in X11 doesn't necessarily copy anything
>>> anywhere until it is pasted,
>> Er, by definition "what paste will insert" is "the contents of the
>> clipboard",
> Er, before you start defining things and presenting examples based on
> your [insult deleted],
What does your classic erroneous presupposition have to do with TeX, Wingate?
We can define "the contents of the clipboard" to be "whatever will be inserted here if I hit the paste key". If doing something changes that, then it clobbers the clipboard. It *doesn't matter*, for this purpose, whether the real storage model is just a pointer and a length. If selecting something else changes what will get pasted, then you've lost what you intended to paste and need to go copy it again, even though "only" a pointer and length got overwritten in memory.
Or to put it more simply: if you select "foo" and cut it, then select "bar", and then paste pastes "bar" instead of overwriting the selected "bar" with "foo", you've just lost "foo". No amount of quibbling about definitions and semantics will change that blunt fact.
And if "foo" was really six paragraphs of your thesis you'd just typed and hadn't saved before cutting it, you are now attached to something by an inclined plane wrapped helically around an axis. Again.
> you should investigate the capabilities and
> limitations of the X11 selection service. It's a different way of doing
> things, with its own advantages and disadvantages. It has other uses than
> manipulating blocks of text.
Are you on crack? The X clipboard ONLY does text, UNLIKE the Windows one which can also do chunks of bitmap and any application-defined data structures you care to name, and which can also store multiple representations, so you can select from a web page, paste in Word and get text with at least some of the italics and other formatting preserved, and paste again in Notepad and get plain text without garbage control characters from Word's .doc format *or* HTML markup (but, no formatting). (And if you *want* the HTML markup, you can select and copy from the browser's "View Source" window instead of the rendered page.)
> Incidentally, if you want to use clipboards in the way you are used to,
> you can save selections to and retrieve them from any of several (eight,
> maybe more) cut buffers.
Sounds like a bunch of extra work, equivalent to the workaround I'd use on such a broken system anyway, to wit, keep a notepad open and always paste promptly to that and then use it again to copy and paste to the destination later.
>> Copy text in open notepad window.
>> Close notepad window.
>> Click to position point in some other window.
>> Paste => Crash! Tried to read from already-freed memory buffer.
> Nope. Typically, no selection any more, so nothing to paste.
Well, that's even sillier, if anything. Albeit safer. Except if you did something like capture the output of a command in a terminal buffer, select, copy, close terminal window, paste, and oops, nothing!, now need to start ALL OVER AGAIN and this time make sure after the copy to click to the other window rather than close the terminal window, paste, then click BACK to the terminal window, and THEN close it ... notice the extra work needed vs. just closing the terminal window and
...
> On 2012-04-13, javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>> On 13/04/2012 11:29 AM, unruh wrote:
>>> I still do not see a charge. I see a statement, which, except for the
>>> attempted pejorative "primative" is simply a statement of fact, not a
>>> charge.
>> In other words, except for some little thing, it's not a charge. In
>> other words, it *is* a charge. Gotcha.
> ?? insults are not charges. They are just insults, especially when they
> are lame insults.
What does your classic erroneous presupposition have to do with TeX, unruh?
The proper role of a CLI is as a kind of REPL for testing and developing scripts, unruh. Using a CLI as the primary manual user interface to the system for other tasks is primitive, unruh. We've had better user interfaces since the sixties, unruh, particularly the Xerox PARC research and the Lisp/Lisp Machine research.
-- public final class JSnarker
extends JComponent
A JSnarker is an NNTP-aware component that asynchronously provides snarky output when the Ego.needsPuncturing() event is fired in cljp.
> On 13/04/2012 9:04 AM, Peter Flynn wrote:
> > What's wrong with mv foo bar baz path/
> > If the filenames are short, it's probably faster than trying to click
> > ctl-click ctl-click to accumulate the files,
> I doubt it, unless they're one or two letter names, and there are > possible length limits to the command line to consider. Also, you can > typo a filename (and the longer the total length of the command line the > proportionately more likely there'll be at least one typo) and the > command will abort half-finished, whereas if you miss with a control > click, you might move a file you wanted to keep and miss a file you > wanted to move, but that's quick to correct and doesn't require redoing > half the *rest* of the files. Just two.
> Just don't let your finger slip off the ctrl key. :)
> And then there's the case where you want to move, say, all the pictures > of Tom's face. It's easy to pick those out with a thumbnail view in > Windows Explorer or something similar, and just control-click them as > you see them and then move them. Contrast Unix:
> * ls
> * my-previewer foo.jpg
> * my-previewer foo2.jpg
> * my-previewer bar.jpg
> * Aha, Tom's face!
> * mv bar.jpg dest/
> * my-previewer quux.jpg
> * my-previewer mumble.jpg
> * mv mumble.jpg dest/
> * my-previewer frotz.jpg
> * Uh oh, next name after frotz.jpg scrolled off the top of the screen
> * ls
> * my-previewer quuuux.jpg
> ...
Dumb. A sane image viewer will take multiple image file names on its
command line:
xv *.jpg &
Then, you just need to build one command line, as you scroll through the
images:
mv bar.jpg mumble.jpg ... dest/
> Macs these days run a disguised version of BSD Unix so that's hardly > surprising.
Not disguised, just a different GUI front end. Under the hood it is in
fact BSD.
-- Robert Heller -- 978-544-6933 / hel...@deepsoft.com
Deepwoods Software -- http://www.deepsoft.com/ () ascii ribbon campaign -- against html e-mail
/\ www.asciiribbon.org -- against proprietary attachments
> At Fri, 13 Apr 2012 17:37:21 -0400 Jail Bush Cronies<do....@ask.me> wrote:
>> And then there's the case where you want to move, say, all the pictures
>> of Tom's face. It's easy to pick those out with a thumbnail view in
>> Windows Explorer or something similar, and just control-click them as
>> you see them and then move them. Contrast Unix:
> A sane image viewer will take multiple image file names on its
> command line:
> xv *.jpg&
> Then, you just need to build one command line, as you scroll through the
> images:
> mv bar.jpg mumble.jpg ... dest/
Even if you managed to do this, that's a lot of typing and alt-tabbing, or at least copying, pasting, and alt-tabbing, compared to an Explorer large-icons or filmstrip view and control-clicking.
Or this little trick, if you have disk space to spare: Copy all the images to a temporary directory and then open the first image in the previewer. Delete it from there if it's not Tom's face or else right-arrow to get to the next image. Repeat until it cycles back to the first image you kept. Then move the remaining files to the real destination. This works if you intended a copy rather than a move from the outset. If you really want a move, you can either do this in place, move the survivors to the destination, and then open the recycle bin, sort by date-deleted, and drag everything you deleted back to where it had been (yes, the recycle bin is that safe and reliable if nothing else is concurrently filling that disk).
>> Macs these days run a disguised version of BSD Unix so that's hardly
>> surprising.
> Not disguised, just a different GUI front end. Under the hood it is in
> fact BSD.
And the word for having replaced the face of something while keeping what it is under the hood unchanged is ... "disguised"!
javax.swing.JSnarker <gharri...@boojum.mit.edu> wrote:
> On 13/04/2012 11:50 AM, John Wingate wrote:
>> javax.swing.JSnarker<gharri...@boojum.mit.edu> wrote:
>>> I was talking about moving multiple files.
>> So was I. Your [insult deleted] is showing.
> What does your classic unsubstantiated and erroneous claim have to do > with TeX, Wingate?
Nothing, of course. You were already far off-topic when I entered this
thread.
Only if typing each character. You snipped my explanation of another way
to construct the line with the help of the mouse.
> Some command prompts have a command length limit as > low as 256 characters, so quadruple the number of files or give them > longer names and you have a problem.
Which CLIs have a 256-character limit? The last one I used that had
anything near that was MS-DOS. (I think its limit was 128 characters.)
> ...
> Copy and paste reduces it to six times the work of the GUI solution, to > wit, for each GUI control click you have a GUI/CLI click, shift-click, > copy, click, paste, space. If words select on double-click (de rigueur > in any decent text editor; I wouldn't think it could be relied on in a > terminal whose GUI is a bolted-on afterthought) then it's a mere five > times as much work: double-click, copy, click, paste, space instead of > control-click.
No copy step. And the click is the paste--but now we are getting to
picking nits.
> (The extra work involved in scrolling, if the source directory's > contents won't fit on one screen, is one wheel flick for every N > selected files, so it's lost in the noise. Besides, an ls output you're > copying from and a "list" view in Explorer will be about equally > compact, and the ls output *can't* be scrolled, you have to rerun ls > with a pattern to exclude the first part of the alphabet.)
Not so sure about the compactness of Explorer list view vs the ls view.
And a long ls output scrolls very nicely with a wheel flick in my xterm.
>> I grant that thumbnails can be useful if the files have unhelpful names;
>> e.g., PICT0001.JPG ... PICT0567.JPG.
> Guess how files tend to be named by digital cameras
Guess where I got my example.
>> you should investigate the capabilities and
>> limitations of the X11 selection service. It's a different way of doing
>> things, with its own advantages and disadvantages. It has other uses than
>> manipulating blocks of text.
> Are you on crack? The X clipboard ONLY does text,
That's true of the cut buffers, which predate the selection service.
> UNLIKE the Windows one > which can also do chunks of bitmap and any application-defined data > structures you care to name, and which can also store multiple > representations, so you can select from a web page, paste in Word and > get text with at least some of the italics and other formatting > preserved, and paste again in Notepad and get plain text without garbage > control characters from Word's .doc format *or* HTML markup (but, no > formatting). (And if you *want* the HTML markup, you can select and copy > from the browser's "View Source" window instead of the rendered page.)
The X selection service likewise does source and destination
context-sensitive C&P.
>> Please stop criticizing what you don't seem to know very well
> Oh, I know it very well indeed. You just find my opinions distasteful.
If you know it very well indeed, please show us that you do. Mostly
you have done quite otherwise.
-- John Wingate Mathematics is the art which teaches
joh...@tds.net one how not to make calculations.
--Oscar Chisini