maps

> Hi > > Is it possible to preserve the order of the keys in a map when the map is > returned?: > > map{"b": 2, "c": 2, "a": 3} > > return > > map { > "a": 3, > "b": 2, > "c": 2 > } > > Thanks! > Giuseppe > >

> On Jul 13, 2018, at 12:57 AM, Giuseppe Celano <celano at informatik.uni-leipzig.de> wrote: > > Hi > > Is it possible to preserve the order of the keys in a map when the map is returned?: > > map{"b": 2, "c": 2, "a": 3} > > return > > map { > "a": 3, > "b": 2, > "c": 2 > } > > Thanks! > Giuseppe >

> I was wondering why to change the original order anyway.

> What is the BaseX rationale for order?

> I was also trying to hack it adding initial numbers in the key names, but it does not work (but it does if key names are only numbers).

> Bottom line: I am trying to "closely" reproduce some Python code involving dictionaries, where original order in dictionary.keys() is now interestingly kept (from Python 3.6, as far as I know). Unfortunately, the computation on the dictionary values assumes a specific order of the keys (which is not even an alphabetical one )...

> Which ordering criteria does this particular dictionary use?

> On Jul 13, 2018, at 11:16 AM, Christian Grün <christian.gruen at gmail.com> wrote: > > Hi Giuseppe, > >> I was wondering why to change the original order anyway. > > As maps are defined to be unordered in the spec, it is perfectly legal > for XQuery processors to change the order of map entries when > optimizing the query and constructing the map, and there is no > guarantee that the “original order”, which may be derived from the > string representation of the query, will be preserved in the compiled > query. For example, in the following query, … > > map:merge((1 to 100000) ! map:entry(., .)) > > …an implementation may decide to add the 10 input maps in parallel, or > to reorder them before adding them to the final map. > >> What is the BaseX rationale for order? > > BaseX uses a hash-based map implementation, which has no notion of > ordering [1]. While the entries of these data structure would yield > have a deterministic order when being serialized, this order will have > no similarities with the order in which the map entries were written > down in the textual query. > >> I was also trying to hack it adding initial numbers in the key names, but it does not work (but it does if key names are only numbers). > > It might look like that ;) The chosen order (for parts of the data > structure) depends on the computed hash code [2]. > >> Bottom line: I am trying to "closely" reproduce some Python code involving dictionaries, where original order in dictionary.keys() is now interestingly kept (from Python 3.6, as far as I know). Unfortunately, the computation on the dictionary values assumes a specific order of the keys (which is not even an alphabetical one )... > > Which ordering criteria does this particular dictionary use? > > If performance is not critical (i.e., if you do not plan to store > millions of items in the map), you could write a simple XQuery wrapper > module that does what you need. If performance is an important factor, > and if you have some (more) spare time, you could dive into the > fascinating world of Leo’s XQuery data structures [3] and adapt the > provided map implementations to your needs… > > Christian > > [1] https://en.wikipedia.org/wiki/Hash_array_mapped_trie > [2] https://github.com/BaseXdb/basex/blob/master/basex-core/src/main/java/org/basex/query/value/item/ANum.java#L115 > [3] https://github.com/LeoWoerteler/xq-modules >

> It is the insertion order. I am just converting this code for pure fun and make some tests. I will definitively have a look at Leo's code as well!