Closed philipstarkey closed 6 years ago
Original comment by Chris Billington (Bitbucket: cbillington, GitHub: chrisjbillington).
I suspect the aboce _ensure_str() might fix it, and if not, is a problem anyway. In front_panel_settings.py, there is code like this:
def check_row(self,row,ct_match,blacs_ct,saved_ct):
# If it has a name
if row[0] != "-":
if ct_match:
# Restore
return 1
else:
# Find if this device is in the connection table
connection = blacs_ct.find_by_name(row[0])
connection2 = saved_ct.find_by_name(row[0])
Where row
is a row if a HDF dataset - and strings it contains are numpy bytestrings. So not only would find_by_name be better if it accepted numpy strings, the check row[0] != "-"
is always resulting in False
since "-"
is a unicode strings and row[0]
is a numpy bytestring.
Fixing these case by case might help, but at some point I might go over front_panel_settings
to wrap the reading of the h5 file in something that results in all unicode strings in objects rather than passing rows of numpy arrays around.
Original report (archived issue) by David Meyer (Bitbucket: dihm, GitHub: dihm).
Running BLACS with Python 3 gives the blanking default values issue seen in Issue #37 with edits to the connection table, excluding adding something new. I suspect it is caused by a similar issue as that of Issue #37 but haven't dug into it.
This issue is not present on our Python 2 setup.