Abstract reader/writer design #80
Conversation
src/h5json/hdf5db.py
Outdated
| stop = start + sel_inter.count[dim] | ||
| slices.append(slice(start, stop, 1)) | ||
| slices = tuple(slices) | ||
| # TBD: needs updating to work in the general case! |
| raise IOError(errno.EINVAL, msg) | ||
| return link_json | ||
|
|
||
| def _addLink(self, grp_id, name, link_json): |
There was a problem hiding this comment.
If _addLink is meant to be an internal function, should it be moved to utils?
|
|
||
| for obj_id in self.db: | ||
| # skip deleted objects | ||
| if self.db[obj_id] is not None: |
There was a problem hiding this comment.
This looks like it will still count objects that have been marked as deleted with "DELETED" and marked dirty, since e.g. deleteLink doesn't appear to remove the link from the database.
| try: | ||
| linkObj = parent.get(link_name, None, False, True) | ||
| linkClass = linkObj.__class__.__name__ | ||
| except TypeError: |
There was a problem hiding this comment.
Is there no way to distinguish a UD link from other potential causes of TypeError here and in _getLink?
| self._flush_time = 0.0 | ||
| self._f = None # h5py file handle | ||
|
|
||
| def _copy_element(self, val, src_dt, tgt_dt, fout=None): |
There was a problem hiding this comment.
I think the _copy_* routines here and in h5py_reader should be renamed to indicate that they're also performing a conversion, and the direction of that conversion. Something like _element_json_to_h5py_rep()/_element_h5py_to_json_rep() would prevent confusion between the two identically named functions that perform opposite operations.
| for title in titles: | ||
| link_json = links_json[title] | ||
| link_class = link_json["class"] | ||
| if "DELETED" in link_json: |
There was a problem hiding this comment.
Given that this also handles deletion, would it be more accurate to call _createObjects() something like _syncObjects()?
hdf5db now supports interaces for reader and/or writer objects.
Currently json and hdf5 (using h5py) are supported.