Added serialization of the trie into a memory buffer.

[KOLANICH]

No description provided.

[KOLANICH]

@thep, @tlwg

[thep]

Symbols in libdatrie.map are versioned. Instead of inserting new symbols to old version block, please create a new version block (DATRIE_0.2.13), deriving from DATRIE_0.2.7, and add new symbols there. For libdatrie.def, simple insertion is fine.

[thep]

I'd prefer function naming to start with target object type, so they look like OOP object methods. Please rename get_trie_serialized_size(). get_alpha_map_serialized_size(), get_tail_serialized_size() to trie_get_serialized_size(), alpha_map_get_serialized_size(), tail_get_serialized_size() respectively.

With this, please also update documentation for trie_get_serialized_size(), as it's not for file creation.

[thep]

Sorry for some comments that might introduce some confusion. TrieChar is supposed to be the type of raw characters used in trie node transitions, as opposed to AlphaChar for the Unicode characters in the real world. TrieChar happens to be unsigned char in the current code, but it's not guaranteed to always be so. For example, there is a request in Issue #1 to increase TrieChar size.

In low-level code like serialization, I think we need to be strict on bytes and use unsigned char where byte is actually meant. So, please revert the TrieChar replacements and use the explicit unsigned char instead. You can assume TrieChar to be of one-byte size for now. We can fix this later if TrieChar enhancement actually happens. Sorry for the confusion.

[KOLANICH]

I think we need to be strict on bytes and use unsigned char

unsigned char is not guaranteed to be a byte. uint8_t is.

[thep]

I think we need to be strict on bytes and use unsigned char

unsigned char is not guaranteed to be a byte. uint8_t is.

Yeah, we can do so to the whole project once we decide to move to C99. However, this reminds me of the similar typedefs in datrie/typedefs.h. We can use uint8 for this purpose, then.

[thep]

Please also add this change set to tests/Makefile.am, to get the test run.

diff --git a/tests/Makefile.am b/tests/Makefile.am
index 1c32dc2..41b55cb 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -7,6 +7,7 @@ TESTS = \
 	test_iterator \
 	test_store-retrieve \
 	test_file \
+	test_serialization \
 	test_nonalpha \
 	test_null_trie \
 	test_term_state \
@@ -19,6 +20,7 @@ check_PROGRAMS = \
 	test_iterator \
 	test_store-retrieve \
 	test_file \
+	test_serialization \
 	test_nonalpha \
 	test_null_trie \
 	test_term_state \
@@ -54,6 +56,12 @@ test_file_SOURCES = \
 	$(NULL)
 test_file_LDADD = $(top_builddir)/datrie/libdatrie.la
 
+test_serialization_SOURCES = \
+	test_serialization.c \
+	utils.c \
+	$(NULL)
+test_serialization_LDADD = $(top_builddir)/datrie/libdatrie.la
+
 test_nonalpha_SOURCES = \
 	test_nonalpha.c \
 	utils.c \

[thep]

Merged. Thank you very much for your contribution!

[KOLANICH]

And thank you for the reviews.