- 10 May, 2016 1 commit
-
-
Matthias Vallentin authored
The function basic_streambuf::in_avail() only returns the number of bytes *currently* available in the get area. Since we're dealing with a theoretically unbounded stream, the get area will be reused continuously, and in_avail() only represents a snapshot in time. Instead, we now check whether we got the amount of bytes we needed. This allows the streambuffer to call underflow() internally to replenish its get area.
-
- 09 May, 2016 1 commit
-
-
Dominik Charousset authored
-
- 06 May, 2016 2 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
- 04 May, 2016 3 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
- 02 May, 2016 1 commit
-
-
Dominik Charousset authored
-
- 29 Apr, 2016 8 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
- 27 Apr, 2016 2 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
- 26 Apr, 2016 1 commit
-
-
Dominik Charousset authored
Relates #444.
-
- 25 Apr, 2016 1 commit
-
-
Dominik Charousset authored
-
- 22 Apr, 2016 10 commits
-
-
Dominik Charousset authored
-
Matthias Vallentin authored
It previously only checked for serialization, not deserialization.
-
Dominik Charousset authored
Also move virtual function calls out of destructors. Relates #454.
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Matthias Vallentin authored
Previously, we enforced that the stream buffer must inherit from std::streambuf, which restricts the character type of stream buffers to char only. With uint8_t buffers being as common, this restriction turns out to be too tight. This change relaxes it.
-
Matthias Vallentin authored
When processing large amounts of data, it is common to store it as std::vector<char> or std::vector<uint8_t>. Previously, such types put heavy stress on the serialization framework, because each character was processed separatedly. This patch changes this behavior such that the entire sequence is processed at once.
-
Matthias Vallentin authored
This optimization pays is a common one in binary serialization protocols. We lifted the implementation directly from VAST's serialization framework. The optimization pays off when a sequence has less than 2^21 elements, which is the common case. Moreover, with this change we treat strings as orindary sequences. Consequently, strings are now binary-compatible with byte vectors, in case that ever turns out to be important.
-
- 21 Apr, 2016 4 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
- 19 Apr, 2016 6 commits
-
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-
Dominik Charousset authored
-