Port of clojure.pprint to Basilisp

[ikappaki]

Hi,

can you please review patch to port clojure.pprint to Basilisp. It addresses #513

The approach taken aimed at minimizing modifications to the original code. There are two commits in this PR, the first is the original code taken from clojure.pprint, the second has all the necessary changes I've made to make it work in Basilisp.

The code spans across seven files loaded into the pprint namespace. I've organized them under basilisp.contrib.pprint to denote that this is a contributed package, rather than something we've implemented from scratch. However, I don't have a specific preference for the location of these files.

Since these files are load'ed at runtime rather than being required or imported, they inevitably cause noticeable delays when the namespace is initially imported, in line with the constraints outlined by #878. Two potential solutions come to mind to necessitate these files at compile time:

1. Transform them into regular namespaces that are required by the primary pprint namespace, as suggested. However, I would then need to figure out a way to integrate these variables into the top-level pprint namespace. I'm not sure there is a super use like form that will not only bring them into the basilisp.pprint namespace but also export them as public variables from there?

2. Merge the files into a single extensive pprint.lpy file.

What do you think?

The pprint tests follow the same approach and also need to be reorganized as above.

The list of changes over the original implementations are

pprint.lpy

• commented out *warn-on-reflection* since it is not applicable in Basilisp *warn-on-reflection* support #881

utilities.lpy

• Fake support for array-map under a hash map (order not preserved), Sorted sets, sorted maps, array maps #416
• STM functions are mocked up around an atom, STM #129
• defstruct support using records, Missing core functions/macros #375
• proxy support, taken from latest draft at Proxies #425
• clojure's private lift-ns and strip-ns fn, taken from Clojure verbatim.
• An interface definition of java.io.Writer, required by pprint.

column-writer.lpy

• Changes are mostly around updating Clojure types to basilisp (e.g. String to py/str).

pprint_base

• Introduce new *print-sort-unordered-colls* var, to order keys in
  map, useful for the testing phase.

cl_format.lpy

• Work around issue that seq cannot be used as a field in a record (renamed to seqn), issue when defrecord field names have the same names as internal method names #758
• Use fractions/Fraction as the type equivalent for clojure.lang.Ratio
• No need for the num fn (boxing numbers in java)
• Create an additional :params-keys entry to store the order of the mock array-map.

dispatch.lpy

• implement support for the new *print-sort-unordered-colls* var, so that prinout map output is predictable for testing
• Commented out agent print out support, Agents #413
• Commented out clojure.lang.IPending print out support (not sure what that is)
• Commented out .isArray support (not sure if this is relevant/supported?)
• Commented out basilisp.lang.interfaces/PersistentQueue dispatch method, since it is not available

print_table.lpy

tests

`test_helper.lpy

• introduce platform-newlines

test_cl_format.lpy

• ~C use pr to print out characters and thus are output as strings rather than characters.
• Commented out Decimal128 tests since they are not support in Python

Thanks

[ikappaki]

I put temporary hack in the pprint tests (test_pprint.lpy, contrib/pprint/test_pretty.lpy and contrib/test_cl_format.lpy) because they may be called individually with coverage run (and as such break), instead of loaded by test_pprint.lpy as a whole. It should be resolved once we decide what to do with the `load'ed files.

[ikappaki]

Hi, I've taken the liberty to concatenate the loaded files in contrib.pprint under pprint.clj and test_pprint.clj. This significantly boost loading performance, and shouldn't affect maintainability because no significant changes are expected, the clj code is quite stable and battle tested for many years now.

It is prime for review.

Thanks