Repository navigation
Header order #40
Description
Header order defined by the order that keys are discovered, rather than the order that they are present in.
E.g.
from collections import OrderedDict
from tabulate import tabulate
test_dict = [
OrderedDict([("a", 1), ("b", 2), ("e", 5)]),
OrderedDict([("a", 1), ("b", 2), ("c", 3), ("d", 4), ("e", 5)]),
]
print(tabulate(
test_dict, headers="keys", tablefmt="simple", missingval="0"
))
gives the output:
a b e c d
--- --- --- --- ---
1 2 5 0 0
1 2 5 3 4
whereas I would prefer it in this order:
a b c d e
--- --- --- --- ---
1 2 0 0 5
1 2 3 4 5
Activity
astanin commentedon Mar 5, 2020
This behavior is by design. Reason 1: It's easier to check the first element of a data set, and immediately have an idea how the data will be printed. Reason 2: If some attributes are optional, it's conventional to display them as the last columns of a table, most important information comes first. Reason 3: Existing logic is easier to understand, implement and define unambiguously (otherwise, how sequences "abc", "edab" and "bad" are supposed to me merged? there is no guarantee that all elements have keys sorted in the same way).
A workaround is to add missing values to the first dictionary. Tabulate has to fill missing values anyway. Keys of the first OrderedDict will allow to define a specific order of columns.
Another workaround is to pre-process data and reorder columns:
There is another important reason not to implement this feature: it will break the documented behavior (pull request #41 is breaking existing tests). At this point we cannot exclude that some users are relying on that.
Considering that an alternative to breaking changes is just a line of code, I would rather keep the library as is.
danja100 commentedon Mar 5, 2020
Fair enough. Thanks for the workarounds, I'll take a look at using these.
(apologies, ignore the latest commit, I realised I hadn't configured the test but hadn't seen your reply)