From Newsgroup: comp.lang.tcl

Consider:

set q 10
se op "<"
for {set i 1} {$i $op $q} {incr i} {puts $i}
--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/21/2025 10:25 AM, saito wrote:

Consider:

set q 10
se op "<"
for {set i 1} {$i $op $q} {incr i} {puts $i}

The answer is yes if you put it in an [expr] :-)
--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On Fri, 21 Mar 2025 10:25:17 -0400
saito <saitology9@gmail.com> wrote:

Consider:

set q 10
set op "<"
for {set i 1} {$i $op $q} {incr i} {puts $i}

Several. Here are two

for {set i 1} "\$i $op \$q" {incr i} {puts $i}

or

for {set i 1} [format {$i %s $q} $op] {incr i} {puts $i}

Regards
--
Emiliano
--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/21/2025 3:34 PM, Emiliano wrote:

On Fri, 21 Mar 2025 10:25:17 -0400

Several. Here are two

for {set i 1} "\$i $op \$q" {incr i} {puts $i}

or

for {set i 1} [format {$i %s $q} $op] {incr i} {puts $i}

Nice solutions, especially the former one. I am surprised that the double-quotes work here, with repeated evaluations. Nice approach.
And [format] looks suspicious in its use here :-)

I myself stumbled upon an approach with [expr] earlier.

--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

saito <saitology9@gmail.com> wrote:

On 3/21/2025 3:34 PM, Emiliano wrote:

On Fri, 21 Mar 2025 10:25:17 -0400

Several. Here are two

for {set i 1} "\$i $op \$q" {incr i} {puts $i}

or

for {set i 1} [format {$i %s $q} $op] {incr i} {puts $i}

Nice solutions, especially the former one. I am surprised that the double-quotes work here, with repeated evaluations. Nice approach.
And [format] looks suspicious in its use here :-)

Remember the rules of Tcl [1]:

[4]
If the first character of a word is double-quote (``"'') then the
word is terminated by the next double-quote character. If
semi-colons, close brackets, or white space characters (including
newlines) appear between the quotes then they are treated as
ordinary characters and included in the word. Command
substitution, variable substitution, and backslash substitution are
performed on the characters between the quotes as described below.
The double-quotes are not retained as part of the word.

When the parser is parsing the command, the third "word" of the first
for loop is:

"\$i $op \$q"

Due to rule four, variable substitution and backslash substution is
performed on this word, because it is double quoted.

So $op is replaced with <, and both \$ are replaced with $

Resulting in |$i < $q| ( | to show edges of the "word").

Then the 'for' command is called, with it's second parameter being
$i < $q

The for loop then calls expr on this parameter, and expr looks up the
value of the i and q variables, and tests them for less than. If the
result is true, the loop continues, if the result is false, the loop
stops.


[1] https://www.tcl-lang.org/man/tcl8.2.3/TclCmd/Tcl.htm
--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/21/2025 6:56 PM, Rich wrote:

When the parser is parsing the command, the third "word" of the first
for loop is:

"\$i $op \$q"

Due to rule four, variable substitution and backslash substution is
performed on this word, because it is double quoted.

I read "double quoted" a few times - I think you meant "double-quote'd".

So since it double-quoted, the substitutions occur as you described. And whatever results from that becomes a "static" argument/parameter to the for-loop. Is this right? I recall warnings against using double quotes
with loops since the substitution occurs once.

I guess I could benefit from this if you explain why the for-loop keeps re-evaluating a double-quoted string.


--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/22/2025 8:44 AM, saito wrote:

On 3/21/2025 6:56 PM, Rich wrote:

When the parser is parsing the command, the third "word" of the first
for loop is:

"\$i $op \$q"

Rich, please ignore my previous post. My eyes simply did not see the backslash-escapes in that double-quoted string. It makes sense now.

--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

saito <saitology9@gmail.com> wrote:

On 3/21/2025 6:56 PM, Rich wrote:

When the parser is parsing the command, the third "word" of the first
for loop is:

"\$i $op \$q"

Due to rule four, variable substitution and backslash substution is
performed on this word, because it is double quoted.

I read "double quoted" a few times - I think you meant "double-quote'd".

So since it double-quoted, the substitutions occur as you described. And whatever results from that becomes a "static" argument/parameter to the for-loop. Is this right?

The argument is always a "static" argument. With double quotes (the
name of ASCII character \ux22) it undergoes a first round of quoting
when the parser is first parsing the command (before the for loop
executes) and then is parsed by "expr" on each iteration of the loop.

I recall warnings against using double quotes with loops since the substitution occurs once.

That depends upon what the initial string in the source is.

This:
set i 0 ; set op < ; set q 5
for {} "$i $op $q" {} {...}

Will have the parser, in the inital parse of the command, replace i, op
and q with "0 < 5" and pass that final string to the for loop, and on
each iterationn of the loop, the loop runs expr on "0 < 5" and yes,
this is a "static" value and the loop runs forever (or until you kill
it).

However, that above was not the original example.

What was in the original example was:
set i 0 ; set op < ; set q 5
for {} "\$i $op \$q" {} {...}

The back slashes in "\$i < \$q" are of critical importance. Their
presence make this **not** a static loop situation. The reason why is
during the first level parse, before the for loop is even run, only one variable is substituted (the op variable). What gets passed to the for command by the parser is "$i < $q", which, when passed to expr, will
look up the current values of i and q to use for the comparison,
resulting in a loop that terminates (assuming the i and q variables are updated by the loop iterations).

I guess I could benefit from this if you explain why the for-loop keeps re-evaluating a double-quoted string.

The for command is defined as always running expr on the middle
argument. Whether you get a loop that looks up variable contents by
that expr call to make the check dynamic, or a loop that runs expr on
the exact same static values for each iteration, depends upon what you
pass to the command. That depends upon what you write that is parsed
by the Tcl parser.



--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 21/03/2025 14:32, saito wrote:

On 3/21/2025 10:25 AM, saito wrote:

Consider:

set q 10
se op "<"
for {set i 1} {$i $op $q} {incr i} {puts $i}

The answer is yes if you put it in an [expr] :-)

The answer is also yes if you use ::tcl::mathop

set q 10
set op "<"
namespace import ::tcl::mathop::$op
for {set i 1} {[$op $i $q]} {incr i} {puts $i}


--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/22/2025 8:11 AM, Rich wrote:

--snip--

The for command is defined as always running expr on the middle
argument. Whether you get a loop that looks up variable contents by
that expr call to make the check dynamic, or a loop that runs expr on
the exact same static values for each iteration, depends upon what you
pass to the command. That depends upon what you write that is parsed
by the Tcl parser.

Actually, the manual for the [for] command does not say it runs [expr], rather, it only says:

Then it repeatedly evaluates test as an expression;

And the command [expr] is not mentioned at all. Also, in the page with the 12 rules, it never defines the word expression.

The [if] command, however, does mention the use of [expr]:

The if command evaluates expr1 as an expression (in the same way that expr evaluates its argument).

I suppose one has to get deep into the weeds and fully understand the algorithm of [expr] to be able to parse it all. There, [expr] does define an expression.

One item that took me forever to understand is why in most commands, words such as in these 3,

set foo bar
set foo {bar}
set foo "bar"

the 2 types of quotes don't change the result here from the unquoted version. But in [expr] and therefore also in the first argument to [if] and the second to [for] a string has to be quoted in one of the 2 ways. So that,

if {$foo eq "bar"} ..
if {$foo eq {bar}} ..

is ok, but

if {$foo eq bar} ..

is not ok. And the reason is that,

expr {$foo eq bar}

also produces an error since operands in [expr] are not the same as tcl words. Here, [expr] complains about a bare word - something I've also not seen defined.

Anyway, there's always something to learn here :)





--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 26/03/2025 04:40, et99 wrote:

On 3/22/2025 8:11 AM, Rich wrote:

--snip--

The for command is defined as always running expr on the middle
argument.  Whether you get a loop that looks up variable contents by
that expr call to make the check dynamic, or a loop that runs expr on
the exact same static values for each iteration, depends upon what you
pass to the command.  That depends upon what you write that is parsed
by the Tcl parser.

Actually, the manual for the [for] command does not say it runs [expr], rather, it only says:

     Then it repeatedly evaluates test as an expression;

And the command [expr] is not mentioned at all. Also, in the page with
the 12 rules, it never defines the word expression.

The [if] command, however, does mention the use of [expr]:

    The if command evaluates expr1 as an expression (in the same way that expr evaluates its argument).

I suppose one has to get deep into the weeds and fully understand the algorithm of [expr] to be able to parse it all. There, [expr] does
define an expression.

One item that took me forever to understand is why in most commands,
words such as in these 3,

   set foo bar
   set foo {bar}
   set foo "bar"

the 2 types of quotes don't change the result here from the unquoted version. But in [expr] and therefore also in the first argument to [if]
and the second to [for] a string has to be quoted in one of the 2 ways.
So that,

    if {$foo eq "bar"} ..
    if {$foo eq {bar}} ..

is ok, but

    if {$foo eq bar} ..

is not ok. And the reason is that,

    expr {$foo eq bar}

also produces an error since operands in [expr] are not the same as tcl words. Here, [expr] complains about a bare word - something I've also
not seen defined.

Anyway, there's always something to learn here :)

FWIW, I remember when I first started to use [expr {$a eq "b"}] if took
me *AGES* to get the syntax right. To be fair to [expr], its error
message (if you omit the double-quotes) is pretty specific, but I still
found it a struggle to get the syntax right.

The lesson I took from this was to stop writing

set i foo

and to start writing

set i "foo"

but the temptation to save a couple of keystrokes is still strong :-)

--- Synchronet 3.20c-Linux NewsLink 1.2

From Newsgroup: comp.lang.tcl

On 3/26/2025 5:35 AM, Alan Grunwald wrote:

FWIW, I remember when I first started to use [expr {$a eq "b"}] if took me *AGES* to get the syntax right. To be fair to [expr], its error message (if you omit the double-quotes) is pretty specific, but I still found it a struggle to get the syntax right.

The lesson I took from this was to stop writing

set i foo

and to start writing

set i "foo"

but the temptation to save a couple of keystrokes is still strong :-)

I too have taken to always use the double quotes, first it better indicates my intent, but also my text editor highlights all "strings" in red so it's much easier to see them. This provides my incentive :)

While I was certainly nitpicking, it does show how difficult it can be to describe a computer language in a language like English. For years I thought it had to be in ""'s because that's how one normally writes strings. Then one day I read [expr] (for the n-teenth time) and finally realized one could also use braces, but one of the two is required. That's when the light-bulb finally glowed for me :)








--- Synchronet 3.20c-Linux NewsLink 1.2