Remove improperly supported "unbounded" size

[greymistcube]

This should follow #124.

Squeezes out another 5% or so from the result of #124. Although technically, unbounded encoding length could be considered as part of the spec, this would be hard to do properly without taking a significant performance hit. If one really wants to support it, the entire stack must be overhauled.

There are two rationales:

• I am skeptical of the original code that it actually supports the feature and works properly it is trying to accommodate.
  • C# itself has a hard limit on an array's length, which is int.MaxValue.
  • That is, one cannot expect byte[] to be of length greater than int.MaxValue. Especially buffer[offset] where offset is greater than int.MaxValue is likely to fail.
• With int.MaxValue, the maximum size of byte[] comes out to be about 2GB. For all intents and purposes, this should be enough for now for encoding a single IValue.
  • Bencode, which forms the basis of the bencodex, originates from encoding torrent metadata files, not the file itself. It was never meant to be big. 🙄

Note that serialization of BigInteger, i.e. an integer without a limit (as long as it doesn't exceed int.MaxValue number of digits in length), is still supported.

Encode() without bloat

Method	SetSize	WordSize	Mean	Error	StdDev	Gen0	Allocated
EncodeList	8	8	3.200 ms	0.3390 ms	0.2243 ms	265.6250	1.6 MB
EncodeDict	8	8	12.088 ms	1.2449 ms	0.8234 ms	453.1250	2.73 MB
EncodeNestedList	8	8	27.358 ms	2.9628 ms	1.9597 ms	1937.5000	11.65 MB
EncodeNestedDict	8	8	121.634 ms	10.6371 ms	5.5634 ms	3500.0000	21.6 MB
EncodeList	8	16	3.773 ms	0.5371 ms	0.3553 ms	382.8125	2.29 MB
EncodeDict	8	16	13.179 ms	0.2856 ms	0.1889 ms	671.8750	4.1 MB
EncodeNestedList	8	16	30.123 ms	4.3146 ms	2.8538 ms	2812.5000	17.14 MB
EncodeNestedDict	8	16	134.295 ms	15.8120 ms	10.4586 ms	5500.0000	33.28 MB
EncodeList	16	8	6.171 ms	0.8664 ms	0.5730 ms	398.4375	2.42 MB
EncodeDict	16	8	25.133 ms	2.0928 ms	1.3842 ms	718.7500	4.33 MB
EncodeNestedList	16	8	95.434 ms	15.1526 ms	10.0225 ms	5833.3333	35.3 MB
EncodeNestedDict	16	8	416.789 ms	55.9215 ms	36.9886 ms	11000.0000	67.73 MB
EncodeList	16	16	6.843 ms	1.0291 ms	0.6807 ms	632.8125	3.8 MB
EncodeDict	16	16	27.280 ms	1.8400 ms	1.0949 ms	1156.2500	7.07 MB
EncodeNestedList	16	16	109.229 ms	17.3151 ms	11.4529 ms	9500.0000	57.27 MB
EncodeNestedDict	16	16	447.549 ms	44.9496 ms	29.7314 ms	18000.0000	113.04 MB

[riemannulus]

Are there any cases where using bencodex needs a long offset in actual data?

[greymistcube]

@riemannulus

Are there any cases where you need a long offset in actual data?

Probably not. As I have mentioned, default byte[] cannot handle long as its indexer, nor can it be longer than int.MaxValue. That is, the original implementation would fail anyway. It would be possible to support long sized byte "array" data, but that would require a complete overhaul. If we decide to do so, as with any kind of support for the data structure that is not natively supported by the language, there will be a huge performance hit.

Also, neither IStore nor IStateStore in libplanet support data types requiring long indexing. So we can neither write or read anyway at this point. 🙄

[riemannulus]

Thanks for your contribution!