diff options
Diffstat (limited to 'tex/context/base/mkxl/syst-ini.mkxl')
-rw-r--r-- | tex/context/base/mkxl/syst-ini.mkxl | 84 |
1 files changed, 77 insertions, 7 deletions
diff --git a/tex/context/base/mkxl/syst-ini.mkxl b/tex/context/base/mkxl/syst-ini.mkxl index c13e5eaef..15e349330 100644 --- a/tex/context/base/mkxl/syst-ini.mkxl +++ b/tex/context/base/mkxl/syst-ini.mkxl @@ -16,13 +16,77 @@ %D this loader is found in \type {syst-tex.tex}. Some of the comment's are Don %D Knuths and more of it can be found in the plain \TEX\ format. %D +%D Successive versions will be adapted to \LUAMETATEX. That also means that we +%D define things a bit differently than in \MKII\ and \MKIV. This is party due to +%D the fact that in this engine we removed some side effects or added features. The +%D original \TEX\ program is well defined and documented and is still the benchmark. +%D The \ETEX\ extensions originally were meant as systematic follow up but that +%D never went beyond a first upgrade. It had a dual mode: compatible or extensible. +%D We always used the second mode. The \LUATEX\ engines don't have that. +%D +%D There are some hard coded properties that relate to for instance node types but +%D we always used abstraction. There are also side effects that (probably) originate +%D in the fact that \ETEX\ had to stay compatible with the dominant set of macros +%D using the original \TEX\ ones (either or not wrapped). Even unwanted side effects +%D eventually become features. In\LUATEX\ and even more in \LUAMETATEX\ we get +%D around that by additional mode variables that you can find below. +%D +%D Side note: those who complain about \TEX\ as rather special language, \ETEX\ +%D extensions that should have been or be different, should really look into how +%D \TEX\ and friends evolved, and not come to quick and wrong conclusions. There are +%D only a very few languages that evolve this way. As a comparison one can wonder +%D why \CCODE\ never had proper strings like \PASCAL\ bit it's just hindsight and +%D discussing it has no use. In a similar fashion one should keep in mind that what +%D you see here is just a logical follow up in \MKII\ where we had to bend the rules +%D and \MKIV\ where we had a bit more available. Not many who look at the latest +%D engines had to jump though all the hoops. +%D +%D This is for instance noticeable in the names of some primitvies: \CONTEXT\ has a +%D concept of protection before \type {\protected} showed up, and similarly had +%D expansion helpers. This is why in \MKII\ you find \type {\protected}, \type +%D {\expanded}, and \type {\unexpanded} commands with different meanings than the +%D later \ETEX\ primitives. In \MKIV\ we stuck to the \CONTEXT\ originals and use +%D \type {\normal...} ones when needed (so \type {\normal...} is there for a +%D reason!). In \LMTX\ we made the transition from \type {\unexpanded} to \type +%D {\protected} so again this file looks different than the \MKIV\ ancestor. +%D +%D When primitives get initialized they come from the \TEX\ namespace (the +%D originals), the \ETEX\ extensions (although we dropped some) and \LUATEX\ which +%D includes the \LUAMETATEX\ ones. The most noticeable \ETEX\ extensions we kept are +%D those that relate to expansion and the expression scanners. The last group has +%D always been somewhat special because (we were told) it was modelled after a macro +%D set so it had to be compatible. It is tempting to replace it (and eventually I +%D might do that) but for now we stick to them (although with an integer divide +%D extension (like the \LUA\ \type {//} and the source has commented binary +%D operators but I haven't yet decided on the symbols to use because we need to +%D avoid expansion issues). For the record: the more extensive, related to \ETEX\ +%D follow up \NTS\ never took of, probably because of the implementation language +%D and because performance made it unuseable. This is also why \LUATEX\ took a +%D different route: extending by an extension language. Of course in the end some +%D core functionality got added too. +%D +%D The most noticeable \LUAMETATEX\ specific code here has to do with additional +%D data types (integer and dimension constants) and macro argument extensions. We +%D also set up some defaults in handling math, languages, etc. Elsewhere in the +%D \CONTEXT\ source the real configuration and usage happens so don't consider this +%D an tutorial (as the plain format is). +%D +%D Category codes are another areas where subtle changed happened. You might not +%D grasp it (maybe because \CONTEXT\ is alien to you) but whenever you see something +%D that looks weird keep in mind that there is a valid reason for it. +%D +%D In due time I will add some more comments here, also because some \CONTEXT\ +%D users might be interested in the hostiry. Now to the code! + %D Characters can have special states, that can be triggered by setting their %D category coded. Some are preset, others are to be set as soon as possible, %D otherwise we cannot define any useful macros. %D %D First we define a bunch of constants. Normally we would \type {\setconstant} %D but we're prestine and have no macros defined yet. Abstraction also makes it -%D possible to avoid the \type {^^} in the input. +%D possible to avoid the \type {^^} in the input. We no longer define the \ETEX\ +%D nodes here (we have not only more, but also different numbers) so we do it +%D elsewhere. % cheatsheet % @@ -311,7 +375,10 @@ %D consecutive ranges more easily, for instance if for \MPLIB\ we want to allocate a %D continuous range of boxes. It also permits us to do a proper upward allocation %D for inserts. The current code evolved from code that dealt with older engines but -%D as all engines now provide many registers we removed all traces. +%D as all engines now provide many registers we removed all traces. If you ever run +%D into the \MKII\ \type {mptopdf} code, you'll notice that some register magic was +%D needed to fit into existing macro packages. The same is true for \PICTEX\ that we +%D used in (gheneric) \PPCHTEX. \ifdefined\writestatus \else %\protected\def\writestatus#1#2{\immediate\write\statuswrite{#1: #2}} @@ -563,8 +630,8 @@ {\showboxdepth \plusthree \showboxbreadth\plusfive} -%D Some expected plain variants follow. We don't reuse registers because we -%D don't want clashes. +%D Some expected plain variants follow. We don't reuse registers because we don't +%D want clashes. These will go away and packages that need them have to define them. \aliased\let\p@ \onepoint \aliased\let\m@ne \minusone @@ -639,9 +706,12 @@ \newif\ifscratchconditionone \newif\ifscratchconditiontwo +%D This used to be a protected definition with a \type {dimexpr} but now we have +%D it natively (less tracking noise). + \aliased\let\htdp\boxtotal -%D A few shortcuts: +%D A few shortcuts (gone by now): % \permanent\protected\def\udef {\protected\def } % \permanent\protected\def\ugdef{\protected\gdef} @@ -660,7 +730,7 @@ \permanent\protected\def\defUmathchar #1#2#3#4{\global\immutable\Umathchardef #1 "#2 "#3 "#4 } %D For a while we keep the following, as systems like tikz need it. Best not use -%D that one \CONTEXT. +%D that one \CONTEXT. It will probably move to the tikz loader. \aliased\let\active\activecatcode @@ -1177,7 +1247,7 @@ {\ifflags#2\permanent\permanent\fi % the rest is taken with the copy \def#1{\syst_suggested_alias#1#2}} -% A few traditional allocations (these might go): +% A few traditional allocations (these might go too): % \countdef \count@ 255 % hm, used in \newif .. todo: replace it there % \dimendef \dimen@ 0 |