Edn parity #3
Edn parity #3
Conversation
Both Clojure and ClojureScript readers, as well as edn.el all allow for this.
The lexer treats each comment line as a separate token, starting with the first semicolon, and up to and including the closing newline. A parser/reducer may choose to merge consecutive comment/whitespace tokens into a single comment AST token.
- parse sets as (edn-set (...)) - parse maps as hash tables - add a clj-parse-edn-read-str that acts like edn-read
| Also added support for comments (seems we didn't have that yet). Tagged literals are still missing, which is one of the main things still need to make the rest of the edn.el test cases pass. |
Have the parse edn functions take an extra argument which is an alist of tagged reader literal handlers.
| Added support for tagged literals. We now support the full edn.el test suite, except for the error cases. |
| (right-char) | ||
| (clj-lex-token :lex-error (buffer-substring-no-properties pos (point)) pos 'error-type :invalid-keyword))))) | ||
| (while (or (clj-lex-symbol-rest? (char-after (point))) | ||
| (equal (char-after (point)) ?#)) |
There was a problem hiding this comment.
Playing around with # I'm noticing it can also be part of symbols and tagged literals:
user=> (let [a#b 1] a#b) 1 user=> (let [#a#b 1] #a#b) RuntimeException No reader function for tag a#b clojure.lang.LispReader$CtorReader.readTagged (LispReader.java:1245)So maybe we could add this equality check to clj-lex-symbol-rest?
There was a problem hiding this comment.
Makes sense as you need that for symbol generation in syntax quotes. I'll add it.
tests/speed-comparison.el Outdated
| @@ -0,0 +1,21 @@ | |||
| (with-current-buffer (find-file-noselect "/home/arne/tmp/edn2.list") | |||
There was a problem hiding this comment.
Maybe include this in the repo, or make it a relative path to some place where devs can set up their own test files locally.
| Just one minor remaining concern, other than that it lgtm =) |
Also includes some .travis.yml changes as a first attempt to get the build going again.
MPL is not compatible with GPL, alas, so we are forced to stick to the GPL. Drop the dependency on `s', add the dependency on `a'
| @plexus I know this is post-morten, but you might wanna also change the license string at the end of the README file. |
| Oops, didn't mean to merge this yet, I guess I pushed to master accidentally. |
2 participants
Several changes to get closer to parity with edn.el.
I copied over the test suite from edn.el and bit by bit made it pass. Some changes I made:
(edn-set (...))4.5e+23):#/#is a valid symbol.)The only thing I didn't copy from edn.el is parsing the
\newline,\tabetc character literals. It parses them as symbols, e.g.'newline, which seems nonsensical. I'm parsing them to numbers like the rest of the character literals (in absence of a character time in elisp). We could also consider parsing them as single character strings, which is what ClojureScript does.I also added a script to compare the speed of edn.el and clj-parse.el, first impression is that clj-parse is about 2 times faster.