awk -v k1=v1... -f src1.awk --source 'CODE' -f src2.awk... 
# I.e. if no -f given then default --source is inferred
awk -v k1=v1... src2.awk...


BEWARE \s does not work!

Character support.
Can specify character sequences like \n or \t or \u000a or \0a?


AIX and Solaris (even nawk) awk doesn't understand regexp matcher \w
SOLARIS awk
    is crippled.
    -v doesn't work.   DOES WORK WITH nawk and xpg4.
    Something else doesn't work, I don't remember what.
        Maybe switch for setting ORS?
        Maybe some regular expression character classes don't work,
        I think the [[:space:]] and similar char classes don't work.
Solaris nawk === regular legacy awk.  /usr/xpg4/bin/awk === regular legacy awk.
IIRC switches to specify inline code and code files differ.
    But data input with no input switches works the same for all variants.
        stdin | awk '...'
        awk '...' file.path

Can freely intermix multiple inline code blocks and source files by using
multiple -f and --source switches in any position.

I believe that these two are always equal (exc. for Solaris awk!):
   awk -v name=val...
   in AWK code:  BEGIN { name = val ...}
So generally, do use -v and don't use var-specific switches like -F and don't
initialize built-in variables in the BEGIN block.

CODE:
    x; # comment to EOL
    /egrep_regexp/|relational_expr|BEGIN|END { action statements}
    [defaults-to-all] [defaults to: { print }]
    Other variants of matcher:  &&, ||, ?, (, !, "start_regexp,stop_regexp"

    function name(params) { statements }
    Must explicitly return if you want a return value
    (i.e. last expression value will not be returned).
    Can only return a scalar, not an assoc. array.
    Parameters by value except that assoc. arrays passed by ref.
    Local variables:  just define extra params and they will behave as locals.

Data Types
No reference variables.
null is just "".  (null == "") is true, so you gain nothing by using 'null'
 and it's misleading.  Just use "".
string  Double-quoted with \-escapes.  Does not do $variable interpolation.
floating-point number
1-dimensional expression-indexed assoc. arrays, aka hashes
    THERE ARE NO LISTS/ARRAYS!
    (x in array)  boolean
    delete array
    delete array[key]
    By just evaluating a reference like this, an element is instantiated:
        if (x[nosuch] == 7)...  ==>   x[nosuch} == ""
Though the indexes are internally strings, and documented as that,
  if you use integral index strings then they are sorted (at least for
  'for x in a' loop) as if native integers.  Integer index values apparently
      With integer keys, Gnu awk always iterates in key order.
      Iteration order is entirely unpredictable with nsh awk.
  get implicitly converted to strings correctly, because this works too!
Like Perl, type is determined by evaluation context.
CONVERSIONS
    To number:  0 + x
    To string:  "" + x  (internally uses CONVFMT and sprintf except integers as integers)
There is some special magic for "numeric strings".  ONLY user-input strings of numbers.
Default values:  0, ""
+= and similar operators do not work with strings.  Must do 'x = x addition'.
Does have ++, etc. operators

Special variables
ARGC
ARGIND
ARGV
RS INPUT record separator.  Single char or regexp.  Default null means blank lines + it adds \n to FS.
RT defaults to what RS matched, Record Terminator
FS obviously, can be set by -F switch also.  Single char or regexp.  Default splits into character fields.  ' ' is magic value meaning \s+.
FIELDWIDTHS breaks into fields by offsets
$0 input record
$1, $2...  input fields
NF  number of input fields
Modifying the input vars causes $0 and deriveds to be regenerated.
CONVFMT (sprintf formatter for numbers)
OFMT  same for output unless integer.  Defaults to %.6g.
OFS  defaults to a space
ORS  [n.b. Output RS === ORS, but Input RS === RS]
ENVIRON array
FILENAME
NR  Recnum (absolute)
FNR  Input recnum within FILENAME
IGNORECASE   Controls all regexp and string ops
PROCINFO array
RSTART, RLENGTH


Action Operators (besides the usuals)
space  string concat
GOTCHA!!!  NOT: ... "str1" + "str2"  or  strV1 + strV2, because concat is:
    "str1" "str2"  or  strV1 strV2
expr ~ /re/, expr !~ /re/  (implies no ^ or $)
next  (like Perl)
nextfile
print [>[>]] appends ORS.  By default prints $0 and ORS.
printf fmt, expr-list [>[>]] file]
retval = system(cmd-line)
Also has | pipe writing

Magic file names
Pseudo /dev/x and network destinations
Reading them gives specialized info.

Slurp entire files (or stdin) into $0:  -v RS='\x04'      or BEGIN{RS="\x04"}

Similar to:  perl -pi -we 's/x/y/'
awk -v RS='\x04' '{ print gensub("x", "y", "g") }'
???   Even \s doesn't work with gawk.  Forget \S!
For example:  awk -v RS='\x04' '{ print gensub(/^\S+\s+/, "", "g") }'
if gensub not available, then similarly:
awk -v RS='\x04' '{ gsub("x", "y"); print; }'
???   Even \s doesn't work with gawk.  Forget \S!
For example:  awk -v RS='\x04' '{ gsub("/\S+\s+/, ""); print; }'

General substitution in a string like perl -we '$x =~ s/x/y/;'
gsub(/x/, y, x);

Win->Unix line delims with gsub:
awk -v RS='\x04' '{ gsub(/\r/, ""); print; }' in > out

sub-string indexing (like for 'substr' and 'index' functions)
is 1-based, not 0-based!

Regexp doesn't understand \d.
Regexp imply no ^ or $, at least for top-level patterns, match functions, ~.

Generally has no switch statement, but gawk can be compiled with it.

Print without trailing newline, like echo -n / echo '\c':
BEGIN { ORS = "" }