be67a29drop wcwidth module in preference to pure-haskell wcwidth implementationAgentM9 years ago dc3ee07fix CSV import/export encoding (now universally UTF-8) all tests passAgentM9 years ago ace7680update codebase for hlint 2.1.1 lots of <$> changesAgentM9 years ago 64dfd80remove extraneous unlock in closeLockFileAgentM9 years ago 56286b0fix new hlint warnings regarding use of fmapAgentM9 years ago c022511compress relvars with gzip, write all atomfuncs to one file references #146AgentM9 years ago 5f6509efix file lock handle leak on connection close System.IO actively prevents opening the same file twice in write mode, so we must resort to a lower-level API luckily, the locking API is self-contained needs Windows mungingAgentM9 years ago a11a49aincrement on-disk format version to account for restoration of IntAtom/IntegerAtom splitAgentM9 years ago 8e05e68bump persistence version due to IntAtom changeAgentM9 years ago b5ef26aincrement db format version due to addition of timestamps improve validation of on-disk database format resolves #133AgentM9 years ago 2aba32aadd commit timestamps the timestamp is from the start of the transaction creation, not when it is committed (a few milliseconds later typically)- this is identical to postgresql's now() stamp resolves #123AgentM9 years ago b27ae9ffix or squelch all hlint hints in src, examples, test directoriesAgentM9 years ago 00c9dc73noch found that automerge was not writing all new transactions to disk this was caused by the transaction graph persistence optimization which assumed writing only the latest heads was safe this has been replaced by commit infrastructure being explicit about which transactions it has added to the graph this is potentially a further optimization since it shouldn't need to check the filesystem for which heads have already been written resolves #128AgentM9 years ago 0704e54switch from cryptonite to lighter-weight cryptohash-sha256 which also offers the benefit of compiling on MacOS 10.7AgentM9 years ago 57faeaeWIP compilation checkpoint when the graph is changed behind the process' back, TransactionIsNotHeadError is returned unconditionally, which is a mistake- it should only return that error only if the same head is being contendedAgentM9 years ago 6ea4ad3functioning atom function and database context function serialization dbc test is brokenAgentM10 years ago f4794d3WIP add dbc function errorsAgentM10 years ago aed8762WIP compilation checkpoint tutd switching to compiling project-m36 library as a dependency for executables switched to Atom ADT AtomFunctionBody somewhat workingAgentM10 years ago