Printable View
Is there any way to do this without going through each character of a string and stripping it down piece by piece?
I'm trying to convert this:
"[[9,0][10,3,["bill",2,1]]]"
into a real array
I could probably use char.at or charAt or whatever it's called if I got desperate, but I'm hoping there's a better, more efficient way.
Any suggestions?
Outputs:Code:var str = "This is a really long sentence.";
var charArray = str.split("");
trace(charArray);
T,h,i,s, ,i,s, ,a, ,r,e,a,l,l,y, ,l,o,n,g, ,s,e,n,t,e,n,c,e,.
perhaps I wasn't clear enough in my intentions
I want to convert that string in my first post into that literal array, as if you'd written it in code without the quotes.
Thanks though.
I suppose you could rinse and repeat until you cleared out all the ['s and ]'s. Is it only numerical data, or is it alpha-numerical? Render once gave me a function to parse xml strings into a real function, but I really hacked it to pieces to make it work. Try looking that up; xml data to arrays or something along those lines. It honestly sounds like a parsing issue, but if it's alpha-numerical data that you are working it, that makes it a bit trickier to pull off.
my question is where did the string come from, and why is it in that format(that of an array, if its because you thought you can make an array from it in one command, well then rewrite the string) and why did you make it that way. As far i know, you can do some indexOf and Split commands until you have it the way you want.
OR make the string less complicated... and then make an array with just split commands. As said before, you can parse the scripts.
for example:
"[[9,0][10,3,["bill",2,1]]]" is the same as
9,0-10,3,"bill"|2|1
first split at '-'
then ','
then '|'
Might be some mistake in there, but you get the idea
there's no easy way (i think) you need to do what IP suggested
go through each character, when you encounter a "[" create a new array, find the indexOf the next "]" grab the string between those points, split up using "," delimiter. Keep going through the string, adding the array into the previous array.
I think you do need to go "through each character of a string and stripping it down piece by piece" - like what mr_malee said, but don't forget to disregard "[", "]" and "," if they appear in a quoted section. eg:
[[9,0][10,3,["bill, fred and that other guy",2,1]]]
It's a bit annoying, but only a few lines. Easiest to do it recursively.
There might also be an escape char for handling a quote within quote.
Brendan [pedantic mode]
Crud. Piece by piece then...
oh well, I'll post what I come up with.
Why not :
- replace "[" by "<array>"
- then replace "]" by "</array>"
- loop through the resulting xml, making an array per array-tag, which is the result of the contained string split by ",".
Only flaw I see is that your strings can't contain comma's. Might have some other flaws, but I haven't really thought it through :)
Opps, Captain_404 I didn't see your point! I just noticed your example is valid AS syntax (except for the missing comma). So this actually works as a line:
and it forms the nested arrays you want. Nice. I see now that that's what you meant.Code:var a:Array = [[9,0],[10,3,["bill",2,1]]];
"Debug -> List variables" outputs:
Variable _level0.a = [object #1, class 'Array'] [
0:[object #2, class 'Array'] [
0:9,
1:0
],
1:[object #3, class 'Array'] [
0:10,
1:3,
2:[object #4, class 'Array'] [
0:"bill",
1:2,
2:1
]
]
]
I was thinking that eval() might do it, but it's only for accessing an existing structure, not building one... I don't think there is a way, except piece by piece... as we said.
It does seem so close, doesn't it? But the parsing code that does it is in the IDE, not in the Flash runtime - they want to keep the d/l small, so why include stuff from the parser/compiler? I suppose it isn't so surprising...
Makes me wish I'd checked back sooner...
Anyway, I came up with a function that works well enough for my purposes (and it is far from perfect, but whatever)
Code:var ry = "[[181,69,84,84,3,0],[272,68,84,84,3,0],[377,219,58,85,9,1],[299,69,34,77,1,1],[352,79,109,41,11,2,[dont play,1]][351,116,109,41,11,2,[with fire,1]]]";
var ry2 = [];
var ia = 0;
var ib = 0;
var ic = 0;
var id = -1;
var l = 0;
var lvl = 0;
for (n = 0; n<ry.length; n++) {
trace("sy: "+sy);
trace("ry2: "+ry2);
trace("char: " + ry.charAt(n));
trace("");
//number
if (int(ry.charAt(n))/int(ry.charAt(n)) == 1 or ry.charAt(n) == "0") {
if (l == "num") {
sy[id] *= 10;
sy[id] += int(ry.charAt(n));
} else {
id++;
sy[id] = int(ry.charAt(n));
l = "num";
}
//end bracket
} else if (ry.charAt(n) == "]") {
l = "not";
lvl--;
if (lvl == 2) {
ry2[ia-1][ib+1] = sy;
id = ib;
}
//begin bracket
} else if (ry.charAt(n) == "[") {
l = "not";
lvl++;
if (lvl == 3) {
ib = id;
}
sy = [];
id = -1;
//letter
} else if (ry.charAt(n) != ",") {
if (l == "let") {
sy[id] += ry.charAt(n);
} else {
id++;
sy[id] = ry.charAt(n);
l = "let";
}
//comma
} else {
l = "not";
if (lvl == 1 and ry.charAt(n+1) != "]") {
if (sy.length > 1) {
ry2[ia] = sy;
} else {
ry2[ia] = sy[0];
}
ia++;
}
if (lvl == 2 and ry.charAt(n+1) == "[") {
ry2[ia] = sy;
ia++;
}
}
}
That's really cool! Difficult to follow, but seems to work well... I like how you integrated the int parsing in a natural way (there's parseInt(), but I think you'd still need to step through it as you have, to find the end of the number).
I'm not sure exactly what you're doing with the ia, ib, ic, and id, but I really like the way you plug the child array into the parent array in:
I think the code is simpler if done recursively; but an iteration implementation as you have done is better for some purposes.Code:ry2[ia-1][ib+1] = sy;
I hope to have more time to study your implementation, and knock out a recursive version.
cheers,
Brendan
Hey I finally got around to redo'ing that parse function Render gave me. I wrote a quick example here. The thing about this is that obviously the data exists dynamically. Tracing the new array (as you can see in the example) still gives you the nasty bracketless output. So I wrote a test at the bottom that draws a grid based on the array and it works fine, thus I know the array exists.
You can either manually type in an array or press "RNDM" until you get an array more to your liking. Then press "GO!"
Convert Strings To Arrays
I tried it (UI looks really cool!).
But this
gave this:Code:[1,2,[3,4],5]
Code:1,2,NaN,NaN,NaN
oh, I see - it's only full multi-dimensional arrays; and only 2D (i think).
Which I think was your purpose, so it's perfect. :-)
PS I like the touch of random examples - it also illustrates what you mean by it. Examples good.
That is correct. It supports only square 2d arrays (x by x, not x by y). Eventually I could probably expand it to work indefinitely for any size array of any dimension, but things like [1,2,[3,4],5] are pretty tricky to work with when you are only dealing with strings.
Here's a recursive parser (finally!).
It handles arbitrary depth eg:
[1,2,[3,4],5]
It doesn't handle quoted strings (but easily added to doValue()).
The basic grammar:Code:a1 = "[1,2,[3,4],5]"
//a2 = " [ 1 , 2 , [ 3 , 4 ] , 5 ] " // also handles space
trace("input: "+a1)
result = parse(a1);
trace("output: "+result);
trace("");
trace("Use ctrl-alt-v to see object structure");
trace('(or "Debug -> List Variables")')
var a:Array;
var i:Number;
var c:String;
// S = _* A
function parse(inputA):Array {
a = inputA;
i = -1;
c = a.charAt(++i);
while (c==' ')
c = a.charAt(++i);
if (c=='[') return doA();
else throw new Error("ERROR #1: "+c+": no [ at "+i);
}
// A = [ ( V (, V)* )? ] (bracket-delimited and comma-separated values)
function doA():Array { // on [.
var array:Array = new Array();
c = a.charAt(++i);
if (c==']') return array; // empty array
else array.push(doValue());
while (true) {
if (c==']') return array;
else if (c==',') ; // go on
else throw new Error("ERROR #2: "+c+": no ] or , at "+i);
c = a.charAt(++i);
array.push(doValue());
}
}
// V = A | digit+
// V = _* (A | digit+) _* (with spaces)
function doValue():Object {
while (c==' ') // spaces before
c = a.charAt(++i);
try {
if (c>='0' && c<='9') {
n = c - '0';
while (true) {
c = a.charAt(++i);
if (c==']') return n; // end of num
else if (c==',') return n; // end of num
else if (c==' ') return n; // end of num
else if (c>='0' && c<='9') n = c-'0' + n*10; //go on
// n*10 + c-'0' doesn't work: c is a string (not char), so + c concatenates
else throw new Error("ERROR #3: "+c+": no ] or , or digit at "+i);
}
} else if (c=='[') {
try return doA();
finally c = a.charAt(++i); // execute this after doA()
}
else throw new Error("ERROR #4: "+c+": no [ or digit at "+i);
} finally while (c==' ') // spaces after
c = a.charAt(++i);
}
Code:A = [ ( V (, V)* )? ]
V = A | digit+
Very nice, that seems to work well.
I was able to hack a bit of code for handling strings also, I'm not completely sure that it works in every case, but in any case that I'd need it works perfectly. It's not set up to handle "s though...or 's for that matter, but that shouldn't be that hard to implement for anyone else who would want to use this...
Code:a1 = "[1,abcdef,[3,4],5]"
//a2 = " [ 1 , 2 , [ 3 , 4 ] , 5 ] " // also handles space
trace("input: "+a1);
freeBob = parse(a1);
trace("output: "+freeBob);
trace("");
trace("Use ctrl-alt-v to see object structure");
trace('(or "Debug -> List Variables")');
var a:Array;
var i:Number;
var c:String;
// S = _* A
function parse(inputA):Array {
a = inputA;
i = -1;
c = a.charAt(++i);
while (c==' ')
c = a.charAt(++i);
if (c=='[') return doA();
else throw new Error("ERROR #1: "+c+": no [ at "+i);
}
// A = [ ( V (, V)* )? ] (bracket-delimited and comma-separated values)
function doA():Array { // on [.
var array:Array = new Array();
c = a.charAt(++i);
if (c==']') return array; // empty array
else array.push(doValue());
while (true) {
if (c==']') return array;
else if (c==',') ; // go on
else throw new Error("ERROR #2: "+c+": no ] or , at "+i);
c = a.charAt(++i);
array.push(doValue());
}
}
// V = A | digit+
// V = _* (A | digit+) _* (with spaces)
function doValue():Object {
while (c==' ') // spaces before
c = a.charAt(++i);
try {
if (c>='0' && c<='9') {
n = c - '0';
while (true) {
c = a.charAt(++i);
if (c==']') return n; // end of num
else if (c==',') return n; // end of num
else if (c==' ') return n; // end of num
else if (c>='0' && c<='9') n = c-'0' + n*10; //go on
// n*10 + c-'0' doesn't work: c is a string (not char), so + c concatenates
else throw new Error("ERROR #3: "+c+": no ] or , or digit at "+i);
}
} else if (c/c != 1 and c != ']' and c!= ',' and c != ' ' and c != '[') {
n = c;
while (true) {
c = a.charAt(++i);
if (c==']') return n; // end of num
else if (c==',') return n; // end of num
else if (c==' ') return n; // end of num
else if (c/c != 1 and c != ']' and c!= ',' and c != ' ') n += c; //go on
// n*10 + c-'0' doesn't work: c is a string (not char), so + c concatenates
else throw new Error("ERROR #3: "+c+": no ] or , or digit at "+i);
}
} else if (c=='[') {
try return doA();
finally c = a.charAt(++i); // execute this after doA()
}
else throw new Error("ERROR #4: "+c+": no [ or digit at "+i);
} finally while (c==' ') // spaces after
c = a.charAt(++i);
}
Thanks :)
It's like you're parsing variable names in source code (which don't have quotes either). Hmmm... Maybe you are? You started with a definition that could be parsed as source code, to produce an array...
Nice! c/c != 1 check for numeric - new to me. I guess it works in Flash because anything else gives undefined (instead of stopping with an error).
You could omit "c/c!=1" from the following line, and this would allow digits in the strings, after the first character. Like you can have in identifiers.
Actually, you could leave out that "if" altogether, and just have "n += c", because the other conditions are caught by the previous if's.Code:else if (c/c != 1 and c != ']' and c!= ',' and c != ' ') n += c; //go on
You could also omit "c/c!=1" from the entrance to your section for strings, because any numeric would already be caught by the previous if. But this is just a finesse, and your code is clearer.
Actually, You could omit conditions from that line, except for the c!='[' (and omit even that, if you make it the last "if", after the one that tests for c=="["). This would enable it to accurately interpret some strings that look very wrong to a human. eg
I like wringing out the last little bit of possibility from code, but I'm not confident that these "advantages" ares beneficial in any way. :DCode:[,,,] interpreted as an array of "," and ",".
[,,]] interpreted as an array of "," and "]".
Hmm, what did I start here. :p That c/c!=1 is brilliant. Why oh why isn't there a string class with a bool function like isDigit or isAlpha like c++ has... anyhow, great work. Blew my cobble app out of the water. :p
hmmm... does this work? maybe more efficient than c/c (as if it mattered!)
if (c-c!=0)
EDIT: no, silly me, as soon as the / or - operator sees it's dealing with a non-numeric string, it will return undefined. It won't actually do any division or subtraction. So same efficiency.
EDIT2: or
if (c*0!=0)
Or, could do:
if (isNaN(parseInt(c)))
Grand Totals:Code:100000 Dividing Comparisons Avg: 0.0464 s
100000 Subtracting Comparisons Avg: 0.0457 s
---
100000 Dividing Comparisons Avg: 0.0449 s
100000 Subtracting Comparisons Avg: 0.0463 s
---
100000 Dividing Comparisons Avg: 0.0527 s
100000 Subtracting Comparisons Avg: 0.046 s
---
100000 Dividing Comparisons Avg: 0.0525 s
100000 Subtracting Comparisons Avg: 0.0455 s
---
100000 Dividing Comparisons Avg: 0.0456 s
100000 Subtracting Comparisons Avg: 0.0465 s
---
100000 Dividing Comparisons Avg: 0.0454 s
100000 Subtracting Comparisons Avg: 0.0452 s
---
100000 Dividing Comparisons Avg: 0.0464 s
100000 Subtracting Comparisons Avg: 0.0457 s
---
100000 Dividing Comparisons Avg: 0.045 s
100000 Subtracting Comparisons Avg: 0.0467 s
---
100000 Dividing Comparisons Avg: 0.0462 s
100000 Subtracting Comparisons Avg: 0.0464 s
---
100000 Dividing Comparisons Avg: 0.0451 s
100000 Subtracting Comparisons Avg: 0.046 s
Looks like dividing is actually faster, but, probably negligible. :)PHP Code:1 Million Dividing Comparisons Avg: 0.04702 s
1 Million Subtracting Comparisons Avg: 0.04618 s
That's the way, do the experiment!
But I bet the variation in times is mostly due to Flash noise... over 10 trials, a difference of (0.04702-0.04618) /( (0.04702+0.04618)/2 ) * 100 is only 1.8%... and given the wideness of the variation in the 10 trials (0.0449 - 0.0527 sec for dividing - almost 16%), I'd guess this isn't significant.
If you run it again, I wonder if the overall averages will differ markedly?
100 trials (or 1000) would smooth out this particular source of randomness. hehehe but I really prefer my computational machinery to be deterministic!
PS: I used google to calc the %. convenient
if (c>-1)
EDIT1: and I realize now that of course c/c or c-c will do some calculation - when c is numeric.Code:check("a");
check("0");
check("9");
function check(c) {
trace(c)
trace(c>-1)
if (c>-1) trace("digit")
else trace("not a digit");
trace("")
}