J Locales

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 141

Explicit Definition isolates names with the Local Copula (=.). Locales isolate names with spelling, possibly expressed, usually implied. My trip may have been more worst case than typical. I wrote three verbs.

The name table_bob_, say, is table in locale bob, while table_z_ is table in locale z. The names table_bob_ and table_z_ are locatives. One locale, the base, has no name. One locale, z, is special: its names are known in all locales unless overridden. Locales are usually populated by script. The application is scripted to the base. Secondary verbs script to named locales. If a file called add is:

table=. 1 : 0
by=. ' '&;@,.@[ ,. ]
over=. ({. ; }.)@":@,
tbl=. 1 : '[by] over x./'
 x. tbl
)

... then 0!:0<'\add' scripts the adverb to the base. This scripts the adverb to locale z:

sc_z_=. 0!:0
sc_z_ <'\add'

I script most of my secondaries without change to locale z. If the first sentence of table is removed, and the last two as well, by, over and tbl become global. That is how I script my verb tracking program, and also the DOS editor I described in Vector 11.3. The former is scripted to locale t and the latter to locale e, in each case to keep names isolated by locale after Explicit (:) is removed.

I happened to have cut defined one way in the base and another way in locale z. Locales are populated by definition as well as script. In the next figure 5!:5@< represents the tracking program in locale t. In the figure but not for real, 5!:5@< is also defined in locale z. The verbs trk and zing are distinct. The first box surprised me. I expected trk'cut' to be <;.1. Then I realized locale t knows nothing about the base, just as the base knows nothing about t. After trk_t_ is plugged into z it can be called trk but it still runs in t. On the other hand zing_z_ runs in locale z and zing runs in the base.

cut=. <;.1
cut_z_=. <;.2
zing_z_=. trk_t_=. 5!:5@<
trk_z_=. trk_t_       NB. plug into z
d=. 1 : '(x.''cut'');(x.''cut__'');(x.''cut_z_'')'
(trk d),(zing_z_ d),:(zing d)
┌────┬────┬────┐
│<;.2│<;.1│<;.2│
├────┼────┼────┤
│<;.2│<;.1│<;.2│
├────┼────┼────┤
│<;.1│<;.1│<;.2│
└────┴────┴────┘

After I saw this I added a verb to trk that makes a locative for any name that is not full (cut__ from cut).

My DOS editor in locale e scripts its result to the locale determined by the extension of the file name. I substituted locale for 0!:0 in the definition and added two verbs. For a calendar program in locale c, say, the first verb makes 'sc_c_' from '\j\s\cal.c', and the second verb is almost:

lo=. ('locale=. '"_ ":@, ]) ; (] ":@, '=: 0!:0'"_)
lo'sc_c_'
┌──────────────┬────────────┐
│locale=. sc_c_│sc_c_=: 0!:0│
└──────────────┴────────────┘

For each Format (":) lo really has Do (".). Do is like Execute in APL. I get a nonce error if sc_c_ is not global (=:).

I made no change to applications. I wrote three verbs that construct locatives for two secondaries not in z. All other locatives appear in a profile script. In an application session I never key a locative or see one. Names have grown up in J. Like the best people, they have lost their innocence but kept their simplicity. Locales are a major improvement.

Richard H. Oates
[address omitted]