Our current HDF5 writer saves strings columns as fixed length, using the maximum length of any string to set the string lengths. This only works when you have seen all possible values of the the string, which is not true if you're appending to an existing HDF5 file.
We should explore ways of at least warning users of this behavior and, ideally, providing ways around it. Ideas include
Our current HDF5 writer saves strings columns as fixed length, using the maximum length of any string to set the string lengths. This only works when you have seen all possible values of the the string, which is not true if you're appending to an existing HDF5 file.
We should explore ways of at least warning users of this behavior and, ideally, providing ways around it. Ideas include