A J approach where,
v is a monadic resource acquisition function (file open, acquire lock) with a
defined inverse whose result we normally don't care about (want the result of
u). v inv should raise an error instead of return an error code. there are 2
options for y argument to v inv :
v inv v y (independent of [x] u v y) NB. open file :. close file as example
or
v inv [x] u v y NB. read file :. write file where u changes file contents.
write file as inverse to read file is dyadic (legal inverse), and so function
is incorrect for purposess of read write duality, but such monadic :. dyadic
inverse functionality is easy to write. An interesting concept about duality,
rather than the "inversality" term that duals in J as used for is: :. (dual
definition) is an alternative to gerunds where a modifier can access (any
defined) the dual as a 2nd parameter included in u or v, where inverses reflect
the core J library, but with user verbs, or some of these locks,
acquire/release is close to invesality, there's no conceptual limit on defining
a duality related to modifer purpose.
u for initial simplicity is dyadic
e is error + cleanup (v inv) function that will pickup error in either u or (v
inv): (13!:12 (13!:8) 13!:11 [ v inv@])
but where e can be invoked on error of v u or v inv.
e =: ( :: (13!:12@[(13!:8)13!:11 ))
up1=: 2 : ' ( v inv(@])e ] u ( :: ( (13!:12@[ (13!:8) 13!:11)@''''@(v inv@]))
)) v e '
and then for u ambivalence
uped =: (@]) up1 ]: : up1
to show the 3 possible error "positions" and no error behaviour
2 +up1(-&> ::] :.((-@<)@(1!:2&2))) 3
_3
|domain error
| 2 +up1(-&> ::] :.((-@<)@(1!:2&2)))3
Error in v inv (-@< after display) display shows that v was executed: _3
2 +up1(- :. (1!:2&2) ) '3'
|rank error
| 2 +up1(- :.(1!:2&2))'3'
Error in v (-) bug somewhere (perhaps even my code) that makes this a rank
error instead of domain error. No display shows v inv did not execute.
2 +up1(< :. (1!:2&2) ) 3
┌─┐
│3│
└─┘
||domain error
| 2 +up1(< :.(1!:2&2))3
Error in u. v inv (display) still executes when u causes error.
2 +up1(- :. (1!:2&2) ) 3
_3
_1
No error. v inv executes return value is x u v y (_1)
+:uped(- :.(1!:2&2))3
_3
_6
On Sunday, May 1, 2022, 07:52:05 a.m. EDT, Hauke Rehr <hauke.r...@uni-jena.de>
wrote:
which seems to have been done before
or came up independently
in D, you have slightly more fine-grained options
<open file>
scope(exit){<close file>}
<actual use of file starts here>
<generate opening xml tag>
scope(success){<generate closing xml tag>}
scope(failure){<some cleanup>}
<contents inside that tag are generated here>
a spontaneous reaction:
In J, I’d like to use generators like
fileworker =. workfile thatone
'exitstatus result' =. verb_using_file&.fileworker arg_of_verb
or
memlocker =. lockmem mutex
'exitstatus result' =. atomic_verb&.memlocker arg_of_verb
[do we even need exitstatus?]
one could even make them generate adverbs
but using &. makes the intent clear
so I’d prefer not adding too much syntactic
sugar – but as always, there are different
tastes and digestive systems (called minds) …
so the argument should be passed and the cleanup
assigned as the obverse, potentially supplementing
the result with an exit status value/error code
would that interfere with your ideas?
Am 01.05.22 um 07:54 schrieb Elijah Stone:
> an idea worth stealing
--
----------------------
mail written using NEO
neo-layout.org
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
