Am 21.09.2011 21:55, schrieb Joseph Wright:
> On 21/09/2011 19:36, Frank Mittelbach wrote:
>> Am 21.09.2011 18:48, schrieb Joseph Wright:
>>> On 21/09/2011 17:34, Bruno Le Floch wrote:
>>>> Context uses it in \keeplinestogether, but I don't see why they
>>>> couldn't have used \clubpenalties instead. I don't understand the
>>>> difference between \interlinepenalties and \clubpenalties as described
>>>> in etex_man:
>>>>
>>>> - the ith interline penalty value is used after line i of the paragraph;
>>>> - the ith club penalty value is used after line i of a partial
>>>> paragraph;
>>>
>>> My reading of this was that the variation comes down to whether there is
>>> any display math about, but in such a case I'm doubtful you'd want to
>>> 'keep together' in any case. As you say, in most realistic circumstances
>>> \clubpenalties seems to be quite usable, as you have to start from line
>>> 1 with \interlinepenalties too.
>>
>> I beg to disagree. The two arrays serve different purposes and in fact
>> \insertpenalties is the one that is more useful (and was the original
>> reason for providing the this additional functionality (the others are
>> more an afterthought to make thing orthogonal).
>>
>> Main use case:
>>
>>    Ensure that after a section x lines are kept together with the section
>>
>> for this use have to use \insertpenalties as you do *not* want to
>> restart counting after a displayed equation, but you just want x-lines
>> once. (in that case you need a mechanism to restore the default status
>> after the first paragraph, as LaTeX today does (or rather attempts as it
>> not always works in TeX) with \clubpenalty for ensuring 2 lines after a
>> heading)
>>
>> Minor use case:
>>
>>    At the beginning of a paragraph (or after a displayed equation as that
>> is visually simiar) you want at least 2 lines but preferably 3) then
>> something like
>>
>>   \clubpenalties = 2 10000 8000
>>
>> That would be a setting you could use for a whole range of paragraphs or
>> even for a whole document.
>>
>> you can think of others also in combination with \insertpenalties
>
> (I'm assuming you mean '\interlinepenalties' throughout.)

I'm sorry, yes of course I meant the latter

>
> I guess this partly depends on your view on display stuff.

don't think so, the display stuff is a bit ugly, granted, but it doesn't 
change the fact that \interlinepenalties refers within a single 
paragraph only to a fixed number of lines counted from the beginning 
while \clubpenalties get restarted after each interrupting display and 
that is (depending on the use case) not necessarily what you are looking 
for.

> For me, I'm
> not sure how it fits into the concept of a 'line'. If I have a paragraph
>
>     text text text text text
>     text text text text text
>          display
>          display
>     text text text text text
>     text text text text text
>
> then how many lines does the display part count?

In TeX the display counts as 3 lines (regardless of the number of real 
lines it would occupy in a grid system). That is just the way it is. It 
makes sense for single line displayed equations but if you have a set of 
equations then that counter clearly ends up being off

> Do we go with a
> grid-typesetting approach and measure its eight, divide by the text lien
> height and use the result? Do we ignore it?

real grid typesetting needs to do something like the above (which is 
what I have done in xor for example). But that is not directly related 
to the \interlinepenalties \clubpenalties question which is about 
holding certain lines together or allowing breakpoints with a certain 
penalty
>
> I guess the other problem I have is that the generalisation
> \interlinepenalty =>  \interlinepenalties seems a bit odd.

it is odd in the sense that \interlinepenalty applies to all lines and 
\interlinepenalties only to the first n lines. While in contrast 
\clubpenalty and \widowpenalty refer only to single lines

> The
> \clubpenalty =>  \clubpenalties is easy to understand, as both are
> related to the start of a paragraph.

not quite \clubpenalties refer to the beginning of every partial 
paragraph. However \clubpenalty only refers to the penalty after the 
first line of a paragraph.

So on the whole the generalization is odd in all cases, but that doesn't 
mean it isn't useful.

  On the other hand,
> \interlinepenalty is a 'floating' idea with no link to the position in a
> paragraph, whereas \interlinepenalties does have to start at the beginning.
>
> I guess my overall concern is that I've not seen \interlinepenalties
> used 'in the wild', whereas \clubpenalties and \widowpenalties do seem
> to be in use. Perhaps that's because I'm not really involved in
> typesetting maths (and so the effects of display environments pass me by).

the fact that people use it incorrectly doesn't mean we should continue 
to do so.

Point is when I want to have at least 3 lines after a heading I should set

<heading>
\interlinepenalties = 3 10000 10000 1000
<first para>
\interlinepenalties = 0

now if that para contains a display I still get only 3 lines fixed to 
the heading. However if i would have used \clubpenalties instead above 
it would have gotten the following situation:

<heading>
\clubpenalties = 3 10000 10000 1000
<1 line>  %no break
<2 line>  %no break
<3 line>
<4 line>
<display formula (counts as three lines>
<8 line>  %no break                       <--- not desired
<9 line>  %no break                       <--- not desired
<10 line>
<11 line>
<12 line>
\clubpenalties = 0


you may of course want one or 2 lines after each display, but that is a 
different use case above you get this only after a heading

frank