@kan \mathrm{foo} is basically a single assignment of \fam to a differnt integer so is essentially instantaneous \text{foo} (in amsmath) leaves math mode using 4 \hbox sets the text size in each to normalsize normalsize scriptsize and scriptsize (having to switch a possibly large number of text font settings for each size, then it typesets all four boxes, in a mathchoice and then picks the box depending on whether the current atom is in text mode or superscript etc.
Anyone else stress-testing @Brunolefloch's 'redundant package' code? I'm trying it on long document: had to switch to LuaTeX for the main run to avoid running out of space. Still analysing the first log after ~30 mins!
@DavidCarlisle My 6 year old machine takes a couple of seconds to process the 17 pg document I'm currently working on. TeX is quite fast on modern hardware. I remember it been quite fast in the mid 90s too, when I first started using it.
@kan so you went with a local installation of auctex?
@FaheemMitha yes but even if it took microseconds rather than seconds, you still see the tracingall output scrolling past (either for real or imaginary) and weep for all those lost cpu cycles
@FaheemMitha I don't mind burning the cycles to do the right thing (you wouldn't believe the amount of work \bm does to get a bold symbol, but using \text{log}as opposed to \log takes farfar more execution steps and ends up ith a worse (and as Joseph said logically wrong) result with poor math spacing
and ad "locality": The other way around is generally more advised: to have all the definitions at one place (in the preamble), seperated ffrom the contents. However, I understand that you prefer something different
Each search result is listed on a separate line. The first character of each line indicates the current state of the package: the most common states are p, meaning that no trace of the package exists on the system, c, meaning that the package was deleted but its configuration files remain on the system, i, meaning that the package is installed, and v, meaning that the package is virtual. The second character indicates the stored action (if any; otherwise a blank space is displayed) to be performed on the
From the man page
@DavidCarlisle Thanks. Hmm. I have that installed. But I'm getting the error: \DeclareMathOperator\step ERROR: Undefined control sequence.
@FaheemMitha texdoc amsmath :-) but actually I suspect you were using step as a variable name rather than an operator so \mathrm{step) is probably what you want
@FaheemMitha If you just want a one-off use without defining a command then it is $\operatorname{step} x$ which is spaced like $\log x$
@FaheemMitha \mathop spacing is mainly for prefix functions like \log \sin which get an automatic thin space in \sin x \mathrm{step} would be a mathord and get no space compare \log\log and \mathrm{log}\mathrm{log}
@FaheemMitha mathitalic is just what you normally get in math, compare $difficult$ (math italic) to $\mathit{difficult}$ text italic in math, for multi-letter identifiers you want \mathit and you get ligatures and kerns as usual mathital has wider sidebearings and (almost) no ligatures so things are tyeset as a product of separate variables rather than a word
@tohecz added that at last minute as I wasn't sure, there are no normal ligatures but there may be some special ones (I know we experimented with encodings such that --> ligatured to \longrightarrow at one stage)
You need to modify the macro \enit@setresumekeys to do one declaration globally. Put this into your preamble after \usepackage{enumitem}:
\makeatletter
\def\enit@setresumekeys #1#2#3{%
\enit@toks \expandafter {\enit@savekeys }%
\xdef \enit@afterlist {#2\def \enit@noexcs {enit@resumekeys@#1}{...
damn, rep cap stopped me at 195 ... /needs green tick
@tohecz while you're there could you fix that colorbox copied from the MWE it has two word spaces from missing % at eol and two \noindent and a \par which are not doing anything in an \hbox :-)
@tohecz However, one can avoid that using a "series"; the first enumerate is \begin{enumerate}[series=mc] (the string is arbitrary); the next one will be \begin{enumerate}[resume*=mc]
@egreg ah I noticed the siunitx cols were over-wide but didn't want to look up how to fix that so I don't begrudge you that tick for a better answer:-)
@tohecz To become Legendary you have to work hard: the list is lockstep, Herbert, DavidCarlisle, GonzaloMedina, Werner, egreg, MartinScharrer (most recent awarded first); I earned it some weeks after Martin.
@PeterGrill It shouldn't be, but of course with a large number of databases the pool operations are performed many times. Probably it would be the same if there were enough registers.
@egreg No I don't think so, there is something else going on there. The failure happens at around 16,000 instantiations in a minute or so, but with your patch it runs for hours... And yes it does complete...
"Jarir bin Abdullah said, Messenger of Allah upon him not God have mercy of ruthless people Sahih Bukhari Book unification door saying the Almighty God Say claimed God or claiming Rahman whatever let Fella Beautiful Names"
I know it is not reliable, but this really doesn't look like IT-related
@egreg Yep, I was going to try what I commented in Joseph's answer as that should be much faster, but don't know if there are downsides of that approach...
@PeterGrill Yes, it's significantly slower: for 1000 DB the times are 5 seconds (traditional) and 10 (pool); for 2000 it's 10 (traditional) and 30 (pool). :(
Background:
I am generating a document which tracks the progress on components of a project consisting of a large number of files. For each of these files, a tiny DB is generated and read via the datatool package. This DB contains a list of the related files and and where they were found, and ...
@PeterGrill what's it using count registers for, they can usually be replaced by chardefs (l3 has (or had) a whole mechanism of fake count registers before etex made that (usually) unnecessary)
the l3 fake count just held the integer in a mathchardef and to do arithemetic stick it in count@ do the sums and then mathchardef the answer back so you pay a bit of time but save the register allocation
You need to modify the macro \enit@setresumekeys to do one declaration globally. Put this into your preamble after \usepackage{enumitem}:
\makeatletter
\def\enit@setresumekeys #1#2#3{%
\enit@toks \expandafter {\enit@savekeys }%
\xdef \enit@afterlist {#2\def \enit@noexcs {enit@resumekeys@#1}{...
Background:
I am generating a document which tracks the progress on components of a project consisting of a large number of files. For each of these files, a tiny DB is generated and read via the datatool package. This DB contains a list of the related files and and where they were found, and ...
