Pascal, this looks really interesting. I also needed a safe do / eval for
my JHS spreadsheet formulas. Your implementation looks similar to what I
had in mind. I'm also interested if there's any security gaps in it.
On Wed, Feb 26, 2014 at 1:45 PM, Pascal Jasmin <godspiral2...@yahoo.ca>wrote:
> Jon is a text data transfer format similar to JSON encoding is simply:
>
> JON =: lr_z_ =: 3 : '5!:5 <''y'''
>
> rest of code:
>
> cocurrent 'jon'
> safe1 =: '''_0123456789+*-<>|;,#ijL{()a[]'
> safe2 =: 'u:';'x:';'}.';'}:';'" ';'":';'! ';'$ ';'= ';'^ ';'^.'
> issafe1d =: safe1 e.~ {. &>
> issafe2d =: safe2 e.~ 2&{. each
> isSafe =: */@:(issafe2d +. issafe1d)@:;:
>
> doSafe =: 0 0&$`".@.isSafe
>
> doSafed =: 3 : 0
> o =. doSafe y
> if. o = i.0 0 do. w=. ;: y
> o =. 'terms not safe1' ; (] #~ -.@:issafe1d) w
> o , 'terms not safe2' ; (] #~ -.@:issafe2d) w return. end.
> o
> )
>
>
> the verbs ending in d are not necessary, but are there for debugging when
> an expression doesn't unencode (JOFF)
>
> JOFF =: doSafe_z_ =: doSafe_jon_
>
>
> The advantage of this data format compared to JSON, is that it can be much
> more concise, and for J, orders of magnitude faster than encoding or
> decoding JSON. Also, J's library for JSON is limited in what it can
> validly parse. JON can encode/decode any valid J noun no matter how deeply
> nested. For short nouns, it can be more space efficient than 3!:1
> encoding. JON also encodes binary data (0 1 255 {a.) along side text data
>
> lr (i.10);( 5 * i.10);(2| i.10);(10 2 $ 3 4);' '
> (i.10);(5*i.10);0 1 0 1 0 1 0 1 0 1;(10 2$3 4 3 4 3 4 3 4 3 4 3 4 3 4 3 4
> 3 4 3 4);15$' '
>
> doSafe lr (i.10);( 5 * i.10);(2| i.10);(2 2 $ 3 4);' '
>
> ┌───────────────────┬───────────────────────────┬───────────────────┬───┬───────────────┐
> │0 1 2 3 4 5 6 7 8 9│0 5 10 15 20 25 30 35 40 45│0 1 0 1 0 1 0 1 0 1│3 4│
> │
> │ │ │ │3 4│
> │
>
> └───────────────────┴───────────────────────────┴───────────────────┴───┴───────────────┘
> doSafe '(i.10);( 5 * i.10);(2| i.10)'
> ┌───────────────────┬───────────────────────────┬───────────────────┐
> │0 1 2 3 4 5 6 7 8 9│0 5 10 15 20 25 30 35 40 45│0 1 0 1 0 1 0 1 0 1│
> └───────────────────┴───────────────────────────┴───────────────────┘
>
> Although its intended to reverse the output of lr, doSafe can process a
> subset of J, and so instead of using lr, it's possible to craft expressions
> that are shorter than what lr produces. doSafe is also useful as a
> server/sandbox execution environment, and sending/receiving records and
> lists as single line statements.
>
> doSafe lr ; , 2{"1 ] 1 (5!:7) < 'doSafed_jon_'
> o=.doSafe yif.o=i.0 0do.w=.;:yo=.'terms not
> safe1';(]#~-.@:issafe1d)wo,'terms not safe2';(]#~-.@:issafe2d)wreturn.end.o
>
> The way doSafe works is that first an expression is put through ;: .
> There is a list of safe 1 character prefixes which includes ' and numbers,
> and a few J verb families. i in this list, whitelists both i. and i:.
> There is also a separate whitelist of 2 character prefixes, that can
> whitelist '" ' and '":' without ".. '= ' without =. or =: This means that
> only quoted text (as data) can be run.
>
> The number of whitelisted content could be expanded. 'NB.' could be
> allowed by creating a safe3 list. Many ommitted conjunctions and verbs
> don't need to be. By adding some defined verbs inside the jon locale,
> those could also be whitelisted. For instance the verbs v0,v1...v9 could
> all be "local variables" set by calling a0...a9.
> with current implementation, this is not allowed ('/' not whitelisted)
>
> doSafe '(+/ % #) 1 2 3 '
>
>
> but
> doSafe '(+: + #) 1 2 3 '
> 5 7 9
>
> multiline processing is also possible:
>
> doSafe lr '3213', LF , '222'
> 3213
> 222
>
> doSafe ;. _2 doSafe lr '3213', LF , '10+20' ,LF NB. if you
> expected multiline expressions
> 3213 30
>
> doSafe ;. _2 doSafe lr '3213', LF , '+\20' ,LF,'23423',LF
> 3213
>
> 0 NB. not allowed in one of the lines
>
> 23423
>
>
> My big question though is have I overlooked any potential unsafe code that
> could be run with doSafe?
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
