Jsonpath – a query language for JSON in Postgres [pdf]
sai.msu.su
sai.msu.su
https://goessner.net/articles/JsonPath/
This submission is for PostgreSQL's "JSON Path", which is very similar to the original JSONPath. I'm surprised, and disappointed, that there is no mention of the original JSONPath, which this is obviously inspired by.
(Though the latter is unlikely, since Lisp and FP have taught me that a bunch of functions feeding into each other will do exactly the same thing quite fine.)
John Clements's packaging of SXPath for Racket might be the most accessible way to first look at it: https://docs.racket-lang.org/sxml/sxpath.html
Then Kiselyov's original writings: http://okmij.org/ftp/Scheme/xml.html#SXPath
Before Web browsers had nice developer tools in them, I made a little aid for crafting SXPath queries, for HTML scraping: https://www.neilvandyke.org/racket/webscraperhelper/
With language libraries or as jp CLI:
http://jmespath.org/libraries.html
Example AWS CLI tricks:
https://gist.github.com/magnetikonline/6a382a4c4412bbb68e33e...
https://opensourceconnections.com/blog/2015/07/27/advanced-a...
It is entirely possible that ISO design is based of Goessner’s work, but unfortunately it is just history at this point. Unfortunately this is what standardization often does - the work becomes attributed to organization, and individual authors are forgotten.
[0] https://modern-sql.com/blog/2017-06/whats-new-in-sql-2016
JSON Pointer is not ubiquitous by any means, but it does got use in diverse APIs. Most recently, JMAP uses it for backreferences.
(The naming of these things is a mess. XPath deals in what you could call queries or selectors, not what I would be willing to call paths; JSONPath adopted XPath’s bad name presumably for marketing reasons, and so what became RFC 6901 presumably went with the name JSON Pointer instead of the obvious name of JSON Path to avoid confusion, as it was coming well after JSONPath.)
It's not like optimizing in C, where the compiler looks at the AST and can actually discover something you wouldn't think of. My experience with PostgreSQL's query planner is that it requires so much fine-tuning it's near unusable and super quirky.
That isn't to say that I don't like PostgreSQL and its JSON capabilities - it's just that the more I use it, the more I wish there was something between it and current NoSQL databases...
you can try it here http://json.jasaws.com/
There's a pretty nice jsonpath implementation that was recently produced at https://github.com/pacifica/python-jsonpath2
I haven't switched to it yet, but I'm keeping an eye on it, to see how long it's supported
Weird to see a proper institution on that domain. The .su itself is quite alive and available for registration. It's mostly used as a “quirky knockoff tld,” in the spirit of .io and such.
Mostly inspired by jsonpath -- IIRC I implemented as much of it which made sense like dotted tree lookup "tree.foo.bar", lambda searching (which I was tempted to name "where"(e.g. "for key, value in tree.foo.bar.where(lambda k,v: k > 100)") but thought better of it), equality and comparison operators and a few other things I can't quite remember off the top of my head.
Doesn't do expression parsing (edit: actually...it does a limited form of expression parsing inherited from the C++ lib) but works for the couple simple projects I needed it for.