I have a huge text file whose size is much larger than the total amount of RAM. For example,

...a01012... [a: 120345...6701019876001110222...
t0**00010101fghi bcdef]jj018024[a:__n*m*...
k0102010102eee/01eee/0101//] ghi [a:01]

I need to transform the input file to the output file as follows:

1. In a block of text between [a: and ] — replace any string that matches the regular expression (01|02){1,} with / (at any distance from [a: and ]);
2. In a block of text which is not between [a: and ] — replace any string that matches the regular expression (01|02){1,} with (at any distance from [a: or ], and regardless of whether they exist).

So for the input file above the expected output is

...a2... [a: 120345...67/98760/11/22...
t0**00/fghi bcdef]jj84[a:__n*m*...
k/eee//eee////] ghi [a:/]

There are four specific caveats to consider:

1. The length of each block of text satisfying property #1 or property #2 may be arbitrarily large, possibly even larger than the total amount of available RAM;
2. The length of a string that matches the regular expression (01|02){1,} may be larger than the total amount of available RAM. That is, there may be a string that matches, for example, (01|02){5234567890}, and it must be replaced with the corresponding character in its entirety;
3. I need a reasonable performance (say, ~1–4 hours for 4GB of text, ~2–8 hours for 8GB, ~3–12 hours for 12GB etc. on a typical user PC with 8GB of RAM);
4. The total amount of free space in data storage device is not larger than 4*S, where S is the size of the input file. That is, 4*S is the maximum size of data that a program can need to be stored in non-volatile memory.

Is it possible to solve the problem with sed, Node.js or Perl? If no, what specific scripting language is practically suitable to solve a problem like this, and how?