man/man1/ssam.1 review comments

[russcox_gmail.com, 2009-07-25 19:22:01]

http://codereview.appspot.com/95076/diff/1/2
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/1/2#newcode4
Line 4: me=ssam
replace $me throughout with ssam.

http://codereview.appspot.com/95076/diff/1/2#newcode9
Line 9: Route all output to stdout.  Ignore -n.  Executes commands from -f''s
new line for each argument

http://codereview.appspot.com/95076/diff/1/2#newcode18
Line 18: # only send home a gripe
unnecessary.  put a note in the 
docs if you want people to send email.

http://codereview.appspot.com/95076/diff/1/2#newcode28
Line 28: # handle (no) input file
This sequence would be simpler as

tmp=/tmp/ssam.tmp.$USER.$pid
cat $* >$tmp

http://codereview.appspot.com/95076/diff/1/2#newcode39
Line 39: # do nothing with n since sam only prints on command
This is a bit odd.  Seems like you'd want to do
something like

{
    if(! ~ $#flagf 0) cat $flagf
    if(! ~ $#flage 0) echo $flage
    echo wq
} | sam -d $tmp >[2]/dev/null

if(! ~ $#flagn 0) cat $tmp

[jdc, 2009-07-30 03:33:44]

Rewrote most actual functionality to incorporate Russ' suggestions.  Will upload
new version.

http://codereview.appspot.com/95076/diff/1/2
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/1/2#newcode4
Line 4: me=ssam
On 2009/07/25 19:22:01, rsc wrote:
> replace $me throughout with ssam.
> 

Done.

http://codereview.appspot.com/95076/diff/1/2#newcode9
Line 9: Route all output to stdout.  Ignore -n.  Executes commands from -f''s
On 2009/07/25 19:22:01, rsc wrote:
> new line for each argument
> 

Done.

http://codereview.appspot.com/95076/diff/1/2#newcode18
Line 18: # only send home a gripe
On 2009/07/25 19:22:01, rsc wrote:
> unnecessary.  put a note in the 
> docs if you want people to send email.
> 

Done.

http://codereview.appspot.com/95076/diff/1/2#newcode28
Line 28: # handle (no) input file
On 2009/07/25 19:22:01, rsc wrote:
> This sequence would be simpler as
> 
> tmp=/tmp/ssam.tmp.$USER.$pid
> cat $* >$tmp
> 

Done.

http://codereview.appspot.com/95076/diff/1/2#newcode39
Line 39: # do nothing with n since sam only prints on command
Your !-n prints the entire file no matter what, after all the lines explicitly
printed. sed's !-n prints all lines, but intersperses them with the
corresponding edited line.  sam only prints on command, unlike sed which by
default prints everything. It's less inportant to me to mimic this sed behavior,
since the sam user can easily add ,p to print everything.  So I will exclude the
-n option completely for now, since I don't think the sam user needs it, and sed
scripts are only weakly portable to sam.

http://codereview.appspot.com/95076/diff/1/2#newcode39
Line 39: # do nothing with n since sam only prints on command
On 2009/07/25 19:22:01, rsc wrote:
> This is a bit odd.  Seems like you'd want to do
> something like
> 
> {
>     if(! ~ $#flagf 0) cat $flagf
>     if(! ~ $#flage 0) echo $flage
>     echo wq
> } | sam -d $tmp >[2]/dev/null
> 
> if(! ~ $#flagn 0) cat $tmp
> 
> 

Done, except flagn line.

[russcox_gmail.com, 2009-08-08 20:38:06]

http://codereview.appspot.com/95076/diff/3002/1003
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/3002/1003#newcode29
Line 29: if(! ~ $#flagn 0) cat $tmp
Why did you delete the -n flag?
I think that's pretty useful.

http://codereview.appspot.com/95076/diff/1005/3003#newcode7
Line 7: flagfmt='h,n,f file,e command,G gripe'
flagfmt='n,e script,f sfile'

(matches sed(1))

http://codereview.appspot.com/95076/diff/1005/3003#newcode8
Line 8: args='[infile]
args='[ file ... ]'
is fine.  the rest belongs in the man page.

http://codereview.appspot.com/95076/diff/1005/3003#newcode14
Line 14: If no input file, take input from standard input.'
s/ || ~$flagh 1//

no need for the explicit -h because typing
ssam -? 
will be treated like an invalid option 
and print the usage anyway

http://codereview.appspot.com/95076/diff/1005/3003#newcode18
Line 18: }
sed has a nice feature where if you don't say -e or -f
then the first argument is the command

sed 10q file.c

for example

to do this here i think you'd do

if(~ $#flage 0 && ~ $#flagf 0) {
  if(~ $#* 0) {
    usage
    exit usage
  }
  flage=$1
  shift
}

http://codereview.appspot.com/95076/diff/1005/3003#newcode23
Line 23: {
looking good.
please use a single tab
instead of two spaces for the indentation
same above in the usage message.

[russcox_gmail.com, 2009-08-15 06:52:23]

i think you're pretty close to done here.
i hope you haven't given up.

[jdc, 2009-08-15 14:51:45]

Not at all!  I've been thinking about what you suggested (especially -n), and
writing another tool (vary) on top of ssam.  It takes a sam script and a file
(data, makefile, rc script), applies the sam script to the file, and either
prints the name of the new file or executes it.

It's kind of a ghetto form of aspect-oriented programming, where you have a
simple base script, and then instead of including all the various possible
control paths with if statements, you use sam scripts to actually edit the
script and create a new one.

[jdc, 2009-08-17 01:12:20]

http://codereview.appspot.com/95076/diff/3002/1003
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/3002/1003#newcode29
Line 29: if(! ~ $#flagn 0) cat $tmp
On 2009/08/08 20:38:06, rsc wrote:
> Why did you delete the -n flag?
> I think that's pretty useful.

I figured out a way to make -n provide value.  If set, I add "echo ,p" just
before "echo wq", so the sam script passed in does not have to print any output
itself.

[jdc, 2009-08-17 01:18:48]

On 2009/08/17 01:12:20, jdc wrote:
> http://codereview.appspot.com/95076/diff/3002/1003
> File bin/ssam (right):
> 
> http://codereview.appspot.com/95076/diff/3002/1003#newcode29
> Line 29: if(! ~ $#flagn 0) cat $tmp
> On 2009/08/08 20:38:06, rsc wrote:
> > Why did you delete the -n flag?
> > I think that's pretty useful.
> 
> I figured out a way to make -n provide value.  If set, I add "echo ,p" just
> before "echo wq", so the sam script passed in does not have to print any
output
> itself.

Maybe I should invert that sense, to make it print with an extra ,p at the end,
unless -n is supplied.  This would make it closer to sed, in which -n means
"print nothing unless specifically asked".

[jdc, 2009-08-18 02:29:06]

Updated ssam with your suggestions.  I use the first-argument-as-command feature
of sed all the time, so thanks much for the code sample and detail about no -f
and -e.

Haven't started a man page for ssam yet.  I have written man pages in the past,
so I'm familiar enough with them to have no excuse.  I think I'll use the man
page for sed as a base, and mash it up with references to sam's command language
rather than sed's.

http://codereview.appspot.com/95076/diff/1005/3003
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/1005/3003#newcode7
Line 7: flagfmt='h,f file,e command'
On 2009/08/08 20:38:06, rsc wrote:
> flagfmt='n,e script,f sfile'
> 
> (matches sed(1))
> 

Done.

http://codereview.appspot.com/95076/diff/1005/3003#newcode8
Line 8: args='[infile]
On 2009/08/08 20:38:06, rsc wrote:
> args='[ file ... ]'
> is fine.  the rest belongs in the man page.
> 

Done.

http://codereview.appspot.com/95076/diff/1005/3003#newcode14
Line 14: if(! ifs=() eval `{getflags $*} || ~ $flagh 1){
On 2009/08/08 20:38:06, rsc wrote:
> s/ || ~$flagh 1//
> 
> no need for the explicit -h because typing
> ssam -? 
> will be treated like an invalid option 
> and print the usage anyway
> 

Done.

http://codereview.appspot.com/95076/diff/1005/3003#newcode18
Line 18: 
On 2009/08/08 20:38:06, rsc wrote:
> sed has a nice feature where if you don't say -e or -f
> then the first argument is the command
> 
> sed 10q file.c
> 
> for example
> 
> to do this here i think you'd do
> 
> if(~ $#flage 0 && ~ $#flagf 0) {
>   if(~ $#* 0) {
>     usage
>     exit usage
>   }
>   flage=$1
>   shift
> }
> 
> 

Done.

http://codereview.appspot.com/95076/diff/1005/3003#newcode23
Line 23: if(! ~ $#flagf 0) cat $flagf
On 2009/08/08 20:38:06, rsc wrote:
> looking good.
> please use a single tab
> instead of two spaces for the indentation
> same above in the usage message.
> 

Done.

[russcox_gmail.com, 2009-08-18 05:48:15]

looks great.  
let's move on to the man page.

http://codereview.appspot.com/95076/diff/5004/5005
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/5004/5005#newcode3
Line 3: # send problems to Jason.Catena@gmail.com
I like the sentiment, but I don't like hard-coding email addresses.
Maybe now that there is an issue tracker it suffices to watch
the issues, which I assume you are doing anyway?

http://codereview.appspot.com/95076/diff/5004/5005#newcode22
Line 22: tmp=/tmp/ssam.tmp.$USER.$pid
in unix it is customary to use $TMPDIR
if(~ $#TMPDIR 0)
    TMPDIR=/tmp
tmp=$TMPDIR/ssam.tmp.$USER.$pid

http://codereview.appspot.com/95076/diff/5004/5005#newcode26
Line 26: if(! ~ $#flagf 0) cat $flagf
not that it should ever happen but i'd put these
in the same order they are in the usage: first flage then flagf.
sed does them in the order they appear on the command line
but getflags gives no way to do that

http://codereview.appspot.com/95076/diff/5004/5005#newcode29
Line 29: echo wq
no need for w;
also no need for q;
delete this line

http://codereview.appspot.com/95076/diff/5004/5005#newcode30
Line 30: } | sam -d $tmp >[2]/dev/null
it may be worth
|[2] egrep -v '^ -\. |^\?changed files' >[1=2]
so that except for the 
 -. /tmp/ssam.*.$USER.$pid
and the 
?changed files
at the end, sam error messages are still shown
on standard error.

[russcox_gmail.com, 2009-08-18 06:32:37]

http://codereview.appspot.com/95076/diff/5004/5005
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/5004/5005#newcode25
Line 25: {
i wonder if this block should start with

echo ,

so that the commands begin executing
with the whole input selected.
that seems more useful than starting at 0
and is easy to override.

[jdc, 2009-08-18 14:43:49]

Did some changes, but have issues with two of them, and an idea for another.

http://codereview.appspot.com/95076/diff/5004/5005
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/5004/5005#newcode3
Line 3: # send problems to Jason.Catena@gmail.com
On 2009/08/18 05:48:15, rsc wrote:
> I like the sentiment, but I don't like hard-coding email addresses.
> Maybe now that there is an issue tracker it suffices to watch
> the issues, which I assume you are doing anyway?

How about this instead?

    # log issues to http://codereview.appspot.com/95076/show

http://codereview.appspot.com/95076/diff/5004/5005#newcode22
Line 22: tmp=/tmp/ssam.tmp.$USER.$pid
On 2009/08/18 05:48:15, rsc wrote:
> in unix it is customary to use $TMPDIR
> if(~ $#TMPDIR 0)
>     TMPDIR=/tmp
> tmp=$TMPDIR/ssam.tmp.$USER.$pid
> 

Done.

http://codereview.appspot.com/95076/diff/5004/5005#newcode25
Line 25: {
On 2009/08/18 06:32:40, rsc wrote:
> i wonder if this block should start with
> 
> echo ,
> 
> so that the commands begin executing
> with the whole input selected.
> that seems more useful than starting at 0
> and is easy to override.

I seem to get an extra print of the entire input file.  Maybe it assumes ","
means "0,$p"?

As intended, the first command does not need to specify its range.  But that
makes the first command different than all the others.  If someone uses this
feature but then changes the order of commands---without adjusting which lines
specify a range---the script will quietly fail.  The old first command will not
run, since it selects no range.

I think it would be better for both these reasons not to put in this change.

http://codereview.appspot.com/95076/diff/5004/5005#newcode26
Line 26: if(! ~ $#flagf 0) cat $flagf
On 2009/08/18 05:48:15, rsc wrote:
> not that it should ever happen but i'd put these
> in the same order they are in the usage: first flage then flagf.
> sed does them in the order they appear on the command line
> but getflags gives no way to do that

Before getflags mangles $* we could do this:

    if(echo $* | grep '.*-e.*-f') orderef=ef 
    if(echo $* | grep '.*-f.*-e') orderef=fe 

I specifically chose to apply the file first, before the command line.  Usually
-e commands are more ephemeral, and often are commands under test that could go
into the file for storage.  So I apply them last, for the use case where they
massage the output of a script file.

http://codereview.appspot.com/95076/diff/5004/5005#newcode29
Line 29: echo wq
On 2009/08/18 05:48:15, rsc wrote:
> no need for w;
> also no need for q;
> delete this line
> 

Done.

http://codereview.appspot.com/95076/diff/5004/5005#newcode30
Line 30: } | sam -d $tmp >[2]/dev/null
On 2009/08/18 05:48:15, rsc wrote:
> it may be worth
> |[2] egrep -v '^ -\. |^\?changed files' >[1=2]
> so that except for the 
>  -. /tmp/ssam.*.$USER.$pid
> and the 
> ?changed files
> at the end, sam error messages are still shown
> on standard error.

Agree with the idea, but this implementation produces this error:

?bad delimiter `s'
?bad delimiter `s'
?bad delimiter `a'

[jdc, 2009-09-10 04:48:43]

description of vary:

Apply a sam script (a file with the suffix ".sam") to a file to create a new
file.  The name of the new file is the name of the old file
catenated with a new suffix.  The new suffix is the name of the sam file without
its path or ".sam" suffix.

-v adds, as the second line in the file, a new line which sets the shell
variable or make macro "variant" to the new suffix.

-x runs the new file.  If the user supplies any command options, vary passes
these along when it executes the new file as a command.

[rsc_swtch, 2009-09-10 05:09:38]

let's leave vary out for now;
it can go into a different issue #

[rsc_swtch, 2009-09-10 05:15:42]

http://codereview.appspot.com/95076/diff/9004/8006
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/9004/8006#newcode40
Line 40: will prepend ',' to the command.
will prefix the command with
.L , .

prepend does not have the meaning you are using here.
it means to ponder something deeply.

http://codereview.appspot.com/95076/diff/9004/8006#newcode43
Line 43: option suppresses the default output: the file in its entirety, after
s/:.*//g and delete next line.

http://codereview.appspot.com/95076/diff/9004/8006#newcode48
Line 48: .B ssam -n '10p' file
ssam -n ,10p file

no need for quotes.
note comma

http://codereview.appspot.com/95076/diff/9004/8006#newcode51
Line 51: .B ssam 's,\en\en,\en,g'
more conventional to use / for separator.

ssam 's/^\n//g'

http://codereview.appspot.com/95076/diff/9004/8006#newcode54
Line 54: .B ssam 's,UNIX,& system,g'
s/UNIX/& system/g

http://codereview.appspot.com/95076/diff/9004/8006#newcode66
Line 66: L. E. McMahon,
delete
the sed(1) reference is enough

http://codereview.appspot.com/95076/diff/9004/8006#newcode73
Line 73: If input is from a pipe, buffering may consume
delete current bug.
add

Ssam consumes all of standard input before
running the script.

[rsc_swtch, 2009-09-10 05:27:17]

Let's make these changes to ssam and the manual and
then get that added.  vary can happen in another CL.

http://codereview.appspot.com/95076/diff/8004/9002
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/8004/9002#newcode18
Line 18: start=,
I think this behavior is a mistake: it shouldn't
matter whether the user said -e or -f or whatever.
Instead, it is possible (though not obvious how)
to start with the entire file as the current selection.

See below.

http://codereview.appspot.com/95076/diff/8004/9002#newcode23
Line 23: if(~ $#TMPDIR 0) TMPDIR=/tmp
please split into two lines

http://codereview.appspot.com/95076/diff/8004/9002#newcode28
Line 28: if(! ~ $#start 0) echo -n $start
Selecting everything without printing it is tricky, but this seems to work:

# select entire file
echo ',{'
echo k
echo '}'
echo '0k'

# run scripts, print
if(! ~ $#flagf 0)
    cat $flagf
if(! ~ $#flage 0)
    echo $flage
if(~ $#flagn 0)
    echo ,

http://codereview.appspot.com/95076/diff/8004/9002#newcode34
Line 34: rm $TMPDIR/ssam.*.$USER.$pid
let's make this rm -f $tmp
(i.e. drop the wildcard)
just to be safe.

[jdc, 2009-09-10 18:53:13]

http://codereview.appspot.com/95076/diff/9004/8006
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/9004/8006#newcode40
Line 40: will prepend ',' to the command.
On 2009/09/10 05:15:42, rsc. wrote:
> prepend does not have the meaning you are using here.
> it means to ponder something deeply.

The first three Google results on the word prepend refer to online dictionary
entries which match precisely my meaning, and use computing references.

> will prefix the command with
> .L , .

If I mention the command first then "prefix" probably fits the flow better, but
I want to focus the sentence on the comma rather than the command.

http://codereview.appspot.com/95076/diff/9004/8006#newcode43
Line 43: option suppresses the default output: the file in its entirety, after
On 2009/09/10 05:15:42, rsc. wrote:
> s/:.*//g and delete next line.

Done, but why not be more specific?  It is just patently obvious?

http://codereview.appspot.com/95076/diff/9004/8006#newcode48
Line 48: .B ssam -n '10p' file
On 2009/09/10 05:15:42, rsc. wrote:
> ssam -n ,10p file
> 
> no need for quotes.

Done.

> note comma

Still wondering about the best way to handle this.

I don't think that ssam should select the entire file when it starts in all
cases.  If it did, then the first command in the sam script would have to know
that it is the first command, and the remaining commands would have to know they
are not the first.  Programs that generate a sam script would have to keep track
of which line was first output.

If I know I only have one command (ie, two or more arguments, but no -e or -f),
then I could generate the comma to save the user a keystroke and preserve a bit
of compatibility with sed.  But there's something in forcing the user to
remember the sam range selector in even the simplest commands, since it's a
significant difference from sed.

http://codereview.appspot.com/95076/diff/9004/8006#newcode51
Line 51: .B ssam 's,\en\en,\en,g'
On 2009/09/10 05:15:42, rsc. wrote:
> more conventional to use / for separator.

Done.

> ssam 's/^\n//g'

Doesn't remove the blank lines.  I think sam's ^ only refers to the beginning of
the file or range.

http://codereview.appspot.com/95076/diff/9004/8006#newcode54
Line 54: .B ssam 's,UNIX,& system,g'
On 2009/09/10 05:15:42, rsc. wrote:
> s/UNIX/& system/g
> 

Done.

http://codereview.appspot.com/95076/diff/9004/8006#newcode66
Line 66: L. E. McMahon,
On 2009/09/10 05:15:42, rsc. wrote:
> delete
> the sed(1) reference is enough
> 

Done.

http://codereview.appspot.com/95076/diff/9004/8006#newcode73
Line 73: If input is from a pipe, buffering may consume
On 2009/09/10 05:15:42, rsc. wrote:
> delete current bug.
> add
> 
> Ssam consumes all of standard input before
> running the script.
> 
> 

Done.

[russcox_gmail.com, 2009-09-10 23:38:19]

http://codereview.appspot.com/95076/diff/9004/8006
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/9004/8006#newcode48
Line 48: .B ssam -n '10p' file
There's a difference between starting the
command with a , and starting with , as
the current value of dot.  If you do it by
smashing a , at the beginning, then 
something like sam 3,5p or even ,p fails mysteriously.
If you set dot before executing the command,
then these work just fine.

I think it is totally fine to say "the script starts
executing with . (dot) set to the entire file contents."
It's a nice convenience and easy to explain.

http://codereview.appspot.com/95076/diff/9009/8015#newcode40
Line 40: will prepend ',' to the command.
I'd rather not argue: it's commonly used but still
not a real word.  The whole point is moot since my
comment on the script changed the semantics here
anyway.

[jdc, 2009-09-12 23:56:31]

I guess the effect of the 'k' lines is to select the entire range, for first
commands without ',', while also allowing the user to provide a range if he
likes?

http://codereview.appspot.com/95076/diff/8004/9002
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/8004/9002#newcode18
Line 18: start=,
On 2009/09/10 05:27:17, rsc wrote:
> I think this behavior is a mistake: it shouldn't
> matter whether the user said -e or -f or whatever.
> Instead, it is possible (though not obvious how)
> to start with the entire file as the current selection.
> 
> See below.
> 

Done.

http://codereview.appspot.com/95076/diff/8004/9002#newcode23
Line 23: if(~ $#TMPDIR 0) TMPDIR=/tmp
On 2009/09/10 05:27:17, rsc wrote:
> please split into two lines
> 

Done.

http://codereview.appspot.com/95076/diff/8004/9002#newcode28
Line 28: if(! ~ $#start 0) echo -n $start
On 2009/09/10 05:27:17, rsc wrote:
> Selecting everything without printing it is tricky, but this seems to work:
> 
> # select entire file
> echo ',{'
> echo k
> echo '}'
> echo '0k'
> 
> # run scripts, print
> if(! ~ $#flagf 0)
>     cat $flagf
> if(! ~ $#flage 0)
>     echo $flage
> if(~ $#flagn 0)
>     echo ,
> 

Done.

http://codereview.appspot.com/95076/diff/8004/9002#newcode34
Line 34: rm $TMPDIR/ssam.*.$USER.$pid
On 2009/09/10 05:27:17, rsc wrote:
> let's make this rm -f $tmp
> (i.e. drop the wildcard)
> just to be safe.
> 

Done.

[rsc_swtch, 2009-09-13 21:43:23]

When you reupload, please be sure to upload both ssam and ssam.1

http://codereview.appspot.com/95076/diff/9013/9014
File bin/ssam (right):

http://codereview.appspot.com/95076/diff/9013/9014#newcode29
Line 29: echo ',{
Please make this 4 different echos
to avoid the weird-looking spacing.

echo ',{'
echo k
echo '}'
echo 0k

or else move it up outside the { }

cat $* >$tmp

selectall=',{
k
}
0k'

    # select entire file
    echo $selectall

[rsc_swtch, 2009-09-14 04:19:55]

http://codereview.appspot.com/95076/diff/10004/9018
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/10004/9018#newcode49
Line 49: .B ssam 's,[^A-Za-z],\n,g' file
Nice example.  Let's use something other than s.
Please use slashes so that the deviation from
convention doesn't distract from the content.

ssam 'y/[a-zA-Z]+/ c/\n/' file

[rsc_swtch, 2009-09-15 16:30:59]

please fix these and upload ssam with it
and then we can apply this.

http://codereview.appspot.com/95076/diff/10008/9029
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/10008/9029#newcode28
Line 28: begins with the entire file selected for the first command.
When the script starts, the entire input is selected.

http://codereview.appspot.com/95076/diff/10008/9029#newcode47
Line 47: Print the first 10 lines of the file.
s/the //

http://codereview.appspot.com/95076/diff/10008/9029#newcode49
Line 49: .B ssam 'y/[a-zA-Z]+/ c/\en/' file
s/file/*.ms/

http://codereview.appspot.com/95076/diff/10008/9029#newcode61
Line 61: .B Count frequency of words in stdin.
Let's get this onto one line to avoid
having to reformat the entire section.

ssam 'y/[a-zA-Z]+/ c/\en/' | grep . | sort | uniq -c

No need for sort -n -- leave sorted by word.

[jdc, 2009-09-15 19:51:28]

http://codereview.appspot.com/95076/diff/10008/9029
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/10008/9029#newcode28
Line 28: begins with the entire file selected for the first command.
On 2009/09/15 16:30:59, rsc wrote:
> When the script starts, the entire input is selected.
> 

Done.

http://codereview.appspot.com/95076/diff/10008/9029#newcode47
Line 47: Print the first 10 lines of the file.
On 2009/09/15 16:30:59, rsc wrote:
> s/the //
> 

Done.

http://codereview.appspot.com/95076/diff/10008/9029#newcode49
Line 49: .B ssam 'y/[a-zA-Z]+/ c/\en/' file
On 2009/09/15 16:30:59, rsc wrote:
> s/file/*.ms/

Done.

Isn't this a little *roff-specific?  I take notes in plain text files, for
example, and use *.tex (latex) and *.nw (noweb) files for documents.

Of all the command man pages, only grep.1 lists *.ms along with the other types
of files searched by -G.

http://codereview.appspot.com/95076/diff/10008/9029#newcode61
Line 61: .B Count frequency of words in stdin.
On 2009/09/15 16:30:59, rsc wrote:
> Let's get this onto one line to avoid
> having to reformat the entire section.
> 
> ssam 'y/[a-zA-Z]+/ c/\en/' | grep . | sort | uniq -c
> 
> No need for sort -n -- leave sorted by word.
> 

Done.

In /man/ssam(1) this gets cut off after the grep .

[rsc_swtch, 2009-09-15 20:13:27]

Are you giving codereview bin/ssam too?
It isn't showing up on the review screen from
codereview.appspot.com, but maybe that is
a bug in the uploader or in the appspot server.

http://codereview.appspot.com/95076/diff/10010/9033
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/10010/9033#newcode61
Line 61: Count frequency of words in stdin.
words read from standard input.

[rsc_swtch, 2009-09-15 21:05:45]

this looks fine but ssam is still missing.
you didn't answer my question: are you 
passing ssam to codereview?

[jdc, 2009-09-15 21:12:05]

Uploaded both files.  I hadn't uploaded bin/ssam for a few since there wasn't a
change.  But since that's not apparent unless you see the change set, I'll
upload both from now on.

How would I take vary out of this patch set?

http://codereview.appspot.com/95076/diff/10010/9033
File man/man1/ssam.1 (right):

http://codereview.appspot.com/95076/diff/10010/9033#newcode61
Line 61: Count frequency of words in stdin.
On 2009/09/15 20:13:27, rsc wrote:
> words read from standard input.
> 

Done.

[rsc_swtch, 2009-09-15 23:08:14]

vary is gone by virtue of your not including
it in the most recent patch that you submitted.

codebot: apply

ssam(1): new command

[old.codebot, 2009-09-15 23:08:30]

# codebot: apply

codebot error: unknown full name for jdc <jason.catena@gmail.com>

jdc, please add a note to this issue with
"codebot: name Your Full Name"
on a line by itself without the quotes.

thanks.
codebot


	Return-Path:
<3Xx6wSgMJB3gnoYospYd.YkiYkZaXkpnoY.ospYd.Yki@apphosting.bounces.google.com>
	Received-SPF: Pass; receiver=holo.morphisms.net; client-ip=209.85.132.188;
	Date: Tue, 15 Sep 2009 23:08:15 +0000
	Subject: Re: man/man1/ssam.1 review comments
	From: rsc@swtch.com
	
	vary is gone by virtue of your not including
	it in the most recent patch that you submitted.
	
	codebot: apply
	
	ssam(1): new command
	
	
	
	http://codereview.appspot.com/95076

[rsc_swtch, 2009-09-15 23:20:33]

codebot: apply

ssam(1): new command

[old.codebot, 2009-09-15 23:22:10]

# codebot: apply

applied.

[jdc, 2009-09-15 23:27:33]

Thanks!  Should I see both files on the commit page?  Only the man page seems
there.

[rsc_swtch, 2009-09-30 18:03:22]

On 2009/09/15 23:27:33, jdc wrote:
> Thanks!  Should I see both files on the commit page?  Only the man page seems
> there.

whoops; fixed