D-Shell
This is generated from a plaintext file, dsh.doc.
Return
Next |End
<![CDATA[
The DSH is a minimal shell based on standard algebraic notation with
infix operators such as A+B. The results of the calculations may easily
be moved into e-mail messages, the World Wide Web and other documents.
The user interface is similar to a pocket calculator with AOS
(Algebraic Operating system) and also to the UNIX shell developed in
the 1960s. There are imitations of many of the classical UNIX commands
such as alias, awk, bc, du, find, grep, history, info, ls, man, rm, tar
and vi. The DSH is also called the D-shell. Like a simple push button
calculator the DSH uses the whole display. The entry of numbers, which
may be thousands of digits long, involves full screen editing for almost
every display.
Example 1:
$ 2+2
4
Example 2:
$ find *.txt search hard disk for files ending with .txt
The idea of submitting whole economies to calculations became
popular in the twentieth century firstly with the Russian Revolution
and later on the the Cold War and the formal divisions between
centrally planned economies and the so called Free World. K.Iverson
added the idea of extending the ordinary arithmetic operations to whole
tables of data all at once. A Globally planned economy was already a
real possibility in the 1960s. Iverson's Algebra ran on IBM computers
for decades. It now runs on Windows, Linux, and Psion. It will run on
Mars or wherever the Empire will site its Capital City in the Galaxy.
D-shell commands are translated to algebraic expressions and
simplified. Sometimes the results of the commands are thrown to the
computer for evaluation. This can be very slow even for simple
calculations; and even after extensive calculations and searching there
may still be no answer. This is the case with prime numbers and
conjectures that require proof or dis-proof.
D-shell commands can be typed after a prompt. On supplied scripts
the prompt is a '$', but this could be changed to any character string.
D-shell commands normally start with a lower case letter. If the
user types a line which is not a D-shell command, then it is taken to
be a D4 expression, as defined in <d4.doc>. All commands are defined in
scripts or ascii files with the extension .d4f. The D-shell is a
postcard sized function in one of the main scripts, test.afn, or
dsh.afn. New commands can be added by use of the alias function 'a',
editing a script, or editing one of the alias lists. Most .d4f files
add commands when they are loaded. Some D-shell commands merely act as
starting points for scripts. Start D4 with command line:-
D4X test.afn v=14 x=gong#gong
This will cause the D4 interpretor to run test.afn. The startup
function (STARTUP) loads another script food.d4f, and immediately does
the D-shell command 'pizza'.
The syntax of a function is:-
name {optional argument list} ARGS in scripts.
Arguments are often UNIX style file types, with the '*' character
meaning any string. Sometimes thay may refer to objects in the
workspace. Several aliases can be used for one function, so LINUX and
WINDOWS users can run the same scripts, although results will be
slightly different. The full list of D-shell commands is stored in an
alias-list. This may be permanently changed by either editing test.afn
or d4s.lst. The current list of D-shell commands can always be seen by
typing 'a'.
Many of the D-Shell commands which generate information take you to
either a menu, or to an 'edit' window. Always use the ESC (escape) key
to exit a menu or the CR (Return) key to select a menu option. Most of
the time mouse clicks have no effect if the mouse cursor is within the
D-Shell window.
If a command gives an edit window then the program is running the
#sed editor, a built-in function to edit text buffers. Help is often
available by pressing F2 or ^B.
The 'vi' editor itself can be customised to be a real 'read-only'
editor. It is merely a D4 script. Many supplied script files such as
hebrew.d4f or farsi.d4f contain localised versions of 'vi'. The
existing version of 'vi' is already customised to follow links in .html
files, or '#include' statements in C/C++ programs. It can also follow
links in many other files by using ^L (Control-L). The UNIX script can
work transparently on compressed files.
'vi' differs very much from the UNIX program of the same name. There
are far less commands and special keys. The ^D key takes you to a one
line version of the D-shell with a '>$' prompt.
The cursor movement keys up, down, left, right and the keys marked
HOME, END, PgUp, PgDN and Del have the expected effect. Typing text
either overwrites existing text, or pushes text to the right, depending
the parity of the number of times the INS (insert key) has been
pressed.
ESC exits the editor, while pressing the Up arrow or CR will take
you to editing text. The command line for the editor is highlighted but
a poorly selected colour scheme may not reveal this.
On linux platforms running X-windows the mouse will work. Two buttons
are used. The left button is equivalent to return (CR), and the right button
acts as escape (ESC). In editor mode a mouse click is meant to follow links
based on the context of the screen. The right button always acts as a
sort of return key. Think of it as navigation via plaintext links.
The D-shell is usually started as an intereactive shell. Unlike the
various shells given with UNIX it tries to get all of it's user input
using full screen mode. It is also meant to be adapted to international
languages including Arabic, Chinese and Hebrew inter alia.
]]>
Next | Back|Top|End
<![CDATA[
Name Argument Comment
! DOS-command Run a DOS command.
< DOS-command Run a DOS command, and edit the output.
awk D4-expression Run within editor. String processing.
a Edit alias list.
cd Directory Change working directory.
cls Clear screen.
cp src dst Copy files src to dst.
cr object-name display object to screen.
cut grab D4.CUT screen, using cursor keys. See 'ww'.
del files delete files. See 'rm'
dir pathname Same as DOS directory.
du c:/ Disk usage by directory.
ed file {/string} Edit file, at first line containg string
ev object-name Edit object
f1 Select DOS console font. .bin format
f2 Select DOS console font. Old .com format
f3 Select DOS console font. LINUX .fnt format
file path/selection Select files for editing
find selection Find files on disk matching selection string.
fns Edit functions in workspace
goto For use in D-shell scripts. Very flakey.
grep string files Search for string in files.
help {/string} Look at d4.doc. same as 'man'.
h Types history list to screen. Try 'ev D4.H'
info Browse UNIX style 'info' files.
ldtar Load tar file archiver
ld d4f-file Reads in a new script.
ls List files. Same as 'dir'.
man Same as help.
md pathname Make a new directory
more filename Don't use. Run d4x rather than d4t.
mv src dst Maps onto DOS move.
phone name Searches phone file database.
pwd Print working directory
quit Exit program
rc Read default alias list.
rerun Delete all names, and restart shell.
rm files Delete files
save Write default alias list
si colour Select console colour. Takes a number 0-255.
subtree path Expands directory structure
tar similar to UNIX tar
vars Edit variables in workspace.
vdel file-selector Delete selected files.
ver Version
vi file {/string} Edit file
ww Same as 'cut'
xg Exchange location of block graphic characters
xoff Turn off pulldown menus on error.
All of these commands are aliases. They are defined in many
different places. The effects of the commands will vary greatly for
different users on a UNIX system. The shell itself is defined in a
script, normally 'test.afn'. Startup processing looks for an alias file
called 'd4s.lst'. This adds further aliases. These alias files may be
filled with operating system specific substitutions so that
WINDOWS-NT/95 and LINUX may be made to look very much the same. This
passes to the user the problems of different flavours for different
releases. Effectively every user will have an individual release by
customising the alias definitions. Many will simply be two or three
letter shortcut combinations with a function and variable list syntax.
The user gets the customised shell required. This can greatly assist
internet viewing by storing common URLS in the alias list which work as
portable bookmarks.
Because the shell is itself a script it can easily be adapted to
include system wide variables such as $INET_BROWSER. There are already
several magic variables including $VA which may be mapped to to block
input-output to any device, and $T which works for streams and
character devices. More exotic magic variables can be put in the C
source code to support new features. There are several for dealing with
the keyboard mappings, and the language interpretor can easily look at
local files. The magic variables act as file and device mappings. $SCR
imitates a memory mapped screen. It's use is $SCR[Z]<-DATA where Z is a
set of pixel positions in (y,x) values and DATA is a set of characters
or numbers representing pixel values. D4, like APL and C is a language
with variable types. Scalars may be integer, character or floating
point. Explicit type conversion functions exist, and they must be used
to get the correct results. To convert from integer to floating point
it is always necessary for a floating point constant to appear
somewhere in a calculation. Arithmetic may be done on character strings
but the results will always remain character strings.
Example:
$ 34 0 z 123456789123456*888777444123
109725609496777417037971456
'34 0 z' is a format specification asking for 34 digits. The result
is not completely accurate because the interpretor converts large
numbers to floating point (sometimes). Full precision may be maintained
by entering the numbers as string literals.
$ "123456789123456"*"888777444123"
109725609496777409188649088
When a command contains an error a message is printed and a pull
down menu bar appears. Select the 'Shell' option to return to command
mode. If the error appeared during the saving of a file, then the
'Windows' option allows editing of the text which could not be saved.
It is possible to write this text to a temporary file by typing ESC,
":w " and a file name. This sort of intereaction expects the user to
write file names in full, usually at the bottom of the screen. There is
little place for form filling beaurocracy when dealing with this type
of software. It is best to avoid file names with embedded spaces. While
many operating systems attempt to support this type of 'user friendly'
rubbish few do it really well; embedded blanks in file names are often
necessary to support Japanese or Chinese names, but with a desktop
environment it is easy to create kana icons to point to these files.
It is possible to change the behaviour on error. There is an alias
called 'xoff' which cancels error handling. At an error the interpretor
awaits keyboard input and treats any line as APL.
Example:
$ xoff
0
$ 6y
ERROR [121]: 6y
2+2
4
The line '6y' gives an error. 'y' is a function, and the context
requires an argument. The meaning of the error number is available from
the include files in the source code distribution.
Next the user types '2+2' and the interpretor gives 4. All APL
interpretors give this result. This is known as calculator mode. The
shell just includes calculations. It ensures that the user can check
whether the system has become so degraded that it needs to restarted.
]]>
Next | Back|Top|End
<![CDATA[
Traditional APL recognised a small number of fairly homogeneous data
types. These were character data, integers and real numbers. D4
represents these in the machine with one, four (or two on 16-bit
version) or eight bytes respectively. Given a symbol "A" the function
#nc"A" gives the name class of the viriable as 0 for undefined and then
1,4,8 etc for the type of object.
A character scalar represents a character with ascii value from 0 to
255. A character vector is any number of characters grouped into a
string. A character matrix is a table, or array of characters with a
certain number of rows and columns.
Atomic data types (char, int, double) are grouped in only two
possible ways. These are scalar, vector, and matrix. There is no
difference between a scalar and a one element vector. The null object
may have any type. The size of the object is the number of elements.
The rank of an object is a vector representing the number of dimensions
of the object. Since objects may only have a maximimum of two
dimensions the rank is always either a single number, or a pair of
numbers. Unlike traditional APL there is no scalar with an empty rank.
Numeric literals are represented as strings of decimal digits with
optional decimal points and a preceding sign written as underscore (_).
Text literals are enclosed in double quotes ("). If the double quote is
itself to be included in a string literal, then it must be doubled.
Traditionally a single point has no dimension, a vector had one
dimension, and arrays had two or more dimensions.
]]>
Next | Back|Top|End
<![CDATA[
These functions are given as single keystrokes. They include +, -, *,
% for ordinary arithmetic. Other function symbols include '/', '\',
'=', '!', '~', '>' ,'<', ',', and a selection of lower case letters.
All lower case letters are reserved for use as functions, although not
all are defined. It should also be noticed that many of the functions
are not defined for all ranges of values. This is quite important for
writing programs. A sensible user is not expected to divide by zero,
and similarly a program should not replicate meaningless data. Most
functions in APL are partial functions: that is, they are not defined
for all posible values of the data. This reflects the most 'State of
the Art' hardware technology, where even fairly ordinary operations of
arithmetic can return 'NaN' or 'not a number'. When a function receives
data that it is not designed for it returns an error code and the
system stops for user input.
]]>
Next | Back|Top|End
<![CDATA[
'r' 'r' stands for 'rho', the traditional function for giving or
setting the rank of an object. It used the greek letter rho in
Iverson's first definition of APL.
Like most APL functions 'rho' has two forms.
N <- r X returns rank of X
Z <- N r X constructor.
Here Z becomes a table of rank N whose elements are taken from the
source X. If X is any object of size zero the result has no meaning,
but otherwise the elements of Z are replicated from those of X. With
text vectors this gives a convenient way of replicating words: thus we
can write 12r"cat" to get the string "catcatcat". 0 r X constructs an
empty vector. Expect DOMAIN ERROR or unspecified silly behaviour if N
does not have sensible values when setting the size and shape of an
array. If N is a single number then Z<- N r X creates a vector with N
elements. If N is the pair {n m} the Z <- N r X creates a matrix with m
rows and n columns. A matrix with just one row looks just like a
vector. It is possible to access subsets of elements of vectors or
matrices by subscripting. The syntax for subscripting is to write a
subscript expression within square brackets. The subscript expression
evaluates to a list of integers. Indexing starts with zero. V[0] is
the first element of a vector V.
Example:
5 12 r "ABCD WXYZ STUV"
ABCD WXYZ ST
UVABCD WXYZ
STUVABCD WXY
Z STUVABCD W
XYZ STUVABCD
Example:
5 5 ri25
0 1 2 3 4
5 6 7 8 9
10 11 12 13 14
15 16 17 18 19
20 21 22 23 24
Example:
X<-"this is text"
rX
12
]]>
Next | Back|Top|End
<![CDATA[
i 'iota'. Iota is a greek letter. In the original version of APL
it became known as 'index generator'. Its function is to generate a
vector of counting numbers. Traditionally these could be set to start
from either zero or one, but D4 simply uses zero as the index origin.
i0 is the empty vector, while i1 is the constant 0 and i2 is the vector
{0 1}. For negative values i has no meaning and large values of the
number N in Z<-iN will test the limits of the computer. Generally
the statemement Z<-iN will have no meaning if N > 500000 or so.
Example:
Z<-i11
0 1 2 3 4 5 6 7 8 9 10
]]>
Next | Back|Top|End
<![CDATA[
In APL the operation for concatenating objects is written with the
comma. Thus 1, 2 gives the vector {1 2}. For ordinary strings and
vectors the results do not seem surprising. The value C<- A, B is a
vector of length (rA) + rB, and the elements are those of A followed by
those of B. If either A or B is of character type and the other operand
is numeric the operation should return domain error. If one oporand is
integer and the other is double the result is also double precision.
The integer values are 'floated'.
Because APL deals with arrays there is the possibility that the
expression C<-A,B also yields a table.
Example:
A<-5 5 r i25 & B<-5 2 r 7 8
C<-A,B
0 1 2 3 4 7 8
5 6 7 8 9 7 8
10 11 12 13 14 7 8
15 16 17 18 19 7 8
20 21 22 23 24 7 8
Example:
A<- 4 4 r "abcdefghijklmnop" & B<-4 2r"| "
]]>
Next | Back|Top|End
<![CDATA[
abcd| abcd
efgh| efgh
ijkl| ijkl
mnop| mnop
It can be seen that the comma (,) function is useful in joining
together matrices. Traditional APL allowed the operation to take an
axis modifier so that the user could write C<-A ,[x] B where x stood
for a dimension along which to join A and B. If x was a fraction then a
new dimension was created.
Since D4 allows just two dimensions there is not so much choice. The
expression A<-B ,: C simply creates a matrix with B over C. The colon
serves as the axis modifier. More generally if A and B are matrices
with sizes rA and rB the C<-A, B requires A and B to have the same
number of rows (rA)[0] = (rB)[0] while C<- A ,:B requires the matrices
to have the same number of columns.
In the case either of the argeuments is a single element vector then
it is stretched to match the appropriate dimension of the other
arguement.
Example:
A<-3 3 r "abcdefghi"
A,"|" A,:"-"
abc| abc
def| def
ghi| ghi
---
]]>
Next | Back|Top|End
<![CDATA[
APL has operations for the re-dimensioning of arrays. These are called
take and drop, and they used to be written with upwards and downwards
pointing arrows. In D4 they are abbreviated to 't' and 'd'. If X is an
array (or vector) with size N then the expression S t X gives a new
array of size S.
]]>
Next | Back|Top|End
<![CDATA[
Default arithmetic works on integers. 'Division' is integer
division. so 5 % 2 is 2 but 5 % 2.0 is 2.5. Numbers are 'floated' by
adding a decimal point. Results of floating point operations often get
to the console in a garbled form. When working extensively with
floating point numbers it is necessary to set a format for the data.
The numeric formatting function is written 'z', and it attempts to
format numeric data. Integer lists are printed as a stream of numbers
separated by blanks, while matrices are formatted so that columns do
not overlap. Negative signs are printed as '_' (underscore), rather
than '-'. This is so that the interpretor can re evaluate the output of
a numeric expression to get the resulting numbers.
Several format options have been tried for floating point numbers.
Field widths such as the pair (w,d) z X or the decimal fraction w.d z X
should give a field width w, with d digits after the decimal point.
Another alternative is a picture format:
V<-3.1415926 2.718281828
" #.####" z V gives 3.1416 2.7183
Because APL is an array processing language the printing of several
objects in one statement would often require the arrangement of objects
in tables. This means that no single glorified formatting function
would meet every possible need. Generally a formatting specification
applies to every element in the array which is the right argument of
'z'.
The formatting function has an inverse function 'x' for execute.
This takes any string as an APL expression and evaluates it.
x"2+2"
4
]]>
Next | Back|Top|End
<![CDATA[
Z <- A #outer f B where f is one of the operators:
'+','*','f','c','=','>','<', and '^' for #and. If A is a vector with m
elements, and B a vector with n elements, then the result Z is an m x n
matrix whose elements z[i, j] are given by: z[i,j] = a[i] f b[j] where
f is one of the functions in the list. Outer products are related to
Cartesian products. The outer product takes a function or operator as
argument besides two data values. The function reduces any pair of
values (x,y) to a single scalar z by means of the formula z <- x f y.
The function reduces the pair x,y to a value in the same set from which
x and y were drawn.
Example: A<-2 3 4 & B <- 3 5 7 11
A #outer + B
5 7 9 13
6 8 10 14
7 9 11 15
A #outer f B
2 2 2 2
3 3 3 3
3 4 4 4
A #outer = B
0 0 0 0
1 0 0 0
0 0 0 0
A #outer ^ B
2 0 2 2
3 1 3 3
0 4 4 0
Example: prime numbers up to N
P<-2,(~A #member A #outer * A)/A<-1+2*1+iN<-400
14 10 rP
2 3 5 7 11 13 17 19 23 29
31 37 41 43 47 53 59 61 67 71
73 79 83 89 97 101 103 107 109 113
127 131 137 139 149 151 157 163 167 173
179 181 191 193 197 199 211 223 227 229
233 239 241 251 257 263 269 271 277 281
283 293 307 311 313 317 331 337 347 349
353 359 367 373 379 383 389 397 401 409
419 421 431 433 439 443 449 457 461 463
467 479 487 491 499 503 509 521 523 541
547 557 563 569 571 577 587 593 599 601
607 613 617 619 631 641 643 647 653 659
661 673 677 683 691 701 709 719 727 733
739 743 751 757 761 769 773 787 797
This example from Education Vector, Vol 10 No 4.
A Note on Primes, Ted Emms.
modend.doc
]]>
Next | Back|Top|End
<![CDATA[
If A and B are two matrices, then the matrix vector pruduct of A and
B is given by the expression Z<-A #inner B. Where X is a square matrix,
the matrix inverse of X (if it exists) is given by the expression Z<-
#domino X. By its matrix algebra APL deals with the set of
endomorphisms of a ring. Because the programmer was lazy, the matrix
algebra functions in integer domains stay integer, even at the expense
of word wrapping. With small numbers such as sets of pixel colour
values and immediate array co-ordinates this does not really matter.
Image edge sharpening algorithms can be performed by integer matrix
multiplication.
Examples:
A<-?5 5 r 10 & B<-?5 5r10
A
8 7 1 3 7
5 4 7 9 6
2 4 8 4 0
6 8 9 1 9
6 0 8 8 3
B
2 1 3 5 2
7 3 0 1 5
1 0 1 5 6
4 4 4 6 3
7 1 7 7 0
A #inner B
127 48 86 119 66 Matrix vector product.
123 59 100 160 99
56 30 30 78 84
144 43 94 152 109
73 41 79 139 84
A<-?3 3r6 Random values 0 to 5
A
5 3 3
3 2 0
4 3 0
" #.###" z #domino A Inverse matrix of A
0.000 3.000 -2.000
0.000 -4.000 3.000
0.333 -1.000 0.333
" #.####" z #domino A+N #outer = N<-i3
0.0000 0.3333 0.0000 Inverse matrix of A+1
0.0000 0.3333 0.0000
0.3333 -0.6667 -0.0000
The #domino function is matrix inversion. For this real numbers are
usually required. JPEG decoding is based on matrix multiplication by
special arrays representing roots of unity, that is to say numbers x
satisfying x^n=1 for some n with a value such as 16 or 32. The
algorithm is based on a frequency analysis of colour changes in the
image, or as mathematicians would say a discrete cosine transform.
]]>
Next | Back|Top|End
<![CDATA[
APL had operators known as encode and decode. These work on numbers
and they can be used to display numbers in bases different to 10. Old
APL used a T-sign and an inverted T-sign for these two functions. D4
uses the keywords #base and #represent. These keywords may be abbreviated
to single letter function codes. Use 'b' for #base and 'n' for
#represent. Integer to hexadecimal conversion provides a good example
of these functions. To convert the numbers 34881 12045 _1 to four
digit hexadecimal numbers use the expression:-
(4 r 16) #represent 34881 12045 _1
8 8 4 1
2 15 0 13
15 15 15 15
The result is a matrix with 3 rows and 4 columns. There is one row
for each integer on the right hand side. When a number requires more than
4 places for its representation the result is truncated. This happens
for nagative values.
To convert a matrix to numbers use the #base function.
Example:
5 5 #base 2 2 r 0 1 2 3
1 13
G<-(4 r 16) #represent 34881 12045 _1
G
8 8 4 1
2 15 0 13
15 15 15 15
(4r16) #base G
34881 12045 65535
The #base function can be explained in terms of the matrix vector
product. If A is an (m,n) matrix with m rows and n columns and B is
a vector with components B[0],B[1]...B[n-1] then the expression
Z<- B #base A gives the same result as
Z<- A #inner *\ 1, 1 #drop 1 #phi B
It is possible to use these operators to write functions to convert
numbers to and from hexadecimal, to decode and encode base64 attachments
for e-mail and to manipulate fixed width font files inter alia.
Example:
Z<-C2HEX STR
|Characters to hex
Z<-"0123456789ABCDEF"[,(2r16) #represent #av STR]
Z<-HEX2N X
|Hexadecimal to numbers
Z<-16 16 b (((rX)%2),2)r"0123456789ABCDEF"iX
$ H<-C2HEX "Have a nice day"
$ H
]]>
Next | Back|Top|End
<![CDATA[
$ HEX2N H
72 97 118 101 32 97 32 110 105 99 101 32 100 97 121
$ #av HEX2N H
Have a nice day
The function '#av' which has no abbreviation is used for data
conversions. Y<-#av X will convert a matrix X to characters or numbers
depending on the original type of X. People who learned BASIC will
recognise #av as an extension of the functions ASC() and CHR$().
#av "cat" is 99 97 116 and #av 99 97 116 is "cat".
The functions for base64 encoding are given.
]]>
Next | Back|Top|End
<![CDATA[
ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/
Z<-B64_ENCODE STR;N;K
|Base 64 encode with endpoints.
K<-(N<-rSTR)#mod 3
STR<-STR,(2-K)/ #av 0 & N<-(N+2*K>0)%3
Z<-,(4r64)n(3r256) b (N,3)r #av STR
Z<-BASE64[Z]
->(K=0)/0
|Correction
Z[(rZ)-i1+3-K]<-"="
Z<-B64_DECODE STR;N;K
|Base 64 decode with endpoint
K<-+/STR="="
N<-(rZ<-BASE64iSTR)%4
Z<-#av ,(3r256)n(4 r 64)b(N,4)rZ
Z<-(-K)dZ
Example of use:
$ H<-B64_ENCODE "Send $1000 immediately"
$ H
U2VuZCAkMTAwMCBpbW1lZGlhdGVseQ==
$ B64_DECODE H
Send $1000 immediately
]]>
Next | Back|Top|End
<![CDATA[
#!/home/d4/cp/d4x
# test input from stdin
#do
0 r INIT;Z<-MASSAGE_STDIN;#quit
Z<-INIT
Z<-EV.KB<-EV.ALX<-"DIE"
Z<-MASSAGE_STDIN;T
| "Massage file"
AA: ->(_1 #equiv T<-#read 0)/XX
(T, #av 10) #write 0
4 #print T
->AA
XX: 0 r #close 0
Z<-#sed ":w out.tmp"
Z<-DIE
("bailing out near line 0",#av 10 13) #write 0
Z<-#quit
]]>
Next | Back|Top|End
<![CDATA[
#!/home/d4/cp/d4x
# print arguement list
# define global variables. The first single symbol line in this
# file is the initialisation function.
INIT
Z<-INIT
"run file name ", $1
|Symbolic links to this file can substitute different functions.
|$1 can be parsed and tested.
"args ",ARGS<-$$[2+i17]
Z<-#quit
Z<-TAB GETOPTS STR;_J;_V;_EXP
|Parse -option {value} arguements
Z<-"" & TAB<-#sstomat "|",TAB & STR<-#deb STR
WHILE 0<r_V<-SHIFT
_J<-,((_V," ") #find TAB)[:0]
IF 0<+/_J; _EXP<-(1+r _V)d _J/:TAB; 0 r #do _EXP; REPEATIF 1
Z<-Z," ",_V
WEND
Z<-1dZ
Z<-SEP SHIFT NN;K;L;Q;X
|Get next token from string named NN and update NN
|Unquote quoted text
SEP<-" " #ifndef "SEP"
X<-"" #ifndef NN<-"STR" #ifndef "NN"
L<-0 & IF Q<-X[0] e "'"""; L<-1+(1dX)iX[0]
K<-L+(LdX)iSEP & Z<-KtX
$NL[NN]<-(~f\SEP=X)/X<-(K+1)dX
IF Q; Z<-1d _1dZ
]]>
Next | Back|Top|End
<![CDATA[
The make files for the interpretor were created by APL scripts back
in 1986. It appears that make files are almost never written by humans.
Instead make files are created by running lists of source code files
through various macro processors. These involve a long shell script
called configure. The Gnugo project consists of not more than a
few dozen source files, but the configure script is 5720 lines long.
Many lines of the configure use 'sed', the stream editor.
"hfiles.sh"
#!/bin/sh
#
# This script looks to see if a directory contains .h files
#
for dir in $@; do
for hfile in $dir/*.h; do
if [ -f $hfile ]; then echo $dir; fi
break
done
done
exit 0
"hfiles.df"
#!/home/d4/cp/d4x
#
# check include files in a list of directories
]]>
| Back|Top|End
<![CDATA[
Z<-HFILES;I;X
I<-2
WHILE<rX<-$$[I++]
(0<r, #files X,"/*.h")/X
WEND
Z<-#quit
In fact the second script is worse than useless. It does not test any
individual file and the expression $$[I++] allows only a limited number
of $$ names: the size of the symbol table. This is currently set to
be a few hundred only. That's easy to fix by looking at the argc variable
for the command line and incorporating this into the original symbol
table.
The next thing to do is to add further symbol tables to reflect the
external environment. These will be the usual pointers to applications
and websites or resource files.
]]>
Back to the Top