You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Refactor of PyTables support to allow multiple table types.
This commit allows for support of multiple table types in a pytables hdf file,
supporting the existing infrastructure in a backwards compatible manner (LegacyTable)
while extending to a slightly modified format to support AppendableTables and future support of WORMTables
AppendableTables are implementations of the current table format with two enhancements:
1) mixed dtype support
2) writing routines in cython for enhanced performance
WORMTables (not implemented - but pretty straightforward)
these tables can support a fixed 'table' (meaning not-appendable), that is searchable via queries
this would have greatly enhanced write performance compared with AppendableTables, and a similar read performance profile
In addition, the tables allow for arbitrary axes to be indexed (e.g. you could save a panel that allows indexing on major_axis,minor_axis AND items),
so all dimensions are queryable (currently only major/minor axes allow this query)
all tests pass (with 1 exception)
a frame table round-trip - we fail on a comparison of a sorted index of the frame vs the index of the table (which is as written), not sure why this should be the case?
0 commit comments