On Wed, May 2, 2018 at 11:06 AM, Denis Kudriashov <dionisi...@gmail.com> wrote:
> Hi. > > Maybe #parseSourceCode would be better name for #parseTree. > I've always found it good advice to avoid using a verb phrase to name something which does not entail some kind of action. #parseSourceCode realy reads like something which would parse the source code. #parseTree also has that effect, except for the lack of a tree to parse. > > 2018-05-02 16:33 GMT+03:00 Marcus Denker <marcus.den...@inria.fr>: > >> >> >> > On 27 Apr 2018, at 21:36, Sean P. DeNigris <s...@clipperadams.com> >> wrote: >> > >> > Marcus Denker-4 wrote >> >> I will add comments… >> > >> > I got confused by this again and created an issue: >> > https://pharo.manuscript.com/f/cases/21806/Document-Differen >> ce-between-ast-and-parseTree >> > >> > And then Peter Uhnak reminded me on Discord about this thread. I'm >> happy to >> > add the comments, but not sure I understand the issue well enough. IIUC >> #ast >> > is cached, but #parseTree is not. What I don't understand is the >> purpose of >> > this difference and when one would use one over the other. >> >> the cached #ast is for one interesting for speed (that is, in situations >> where you ask for it often). >> >> The other use-case is if you want to annotate the AST and keep that >> annotation around (till the next >> image save, but you can subscribe to ASTCacheReset and re-install the AST >> in the cache after cleaning. >> (This is used by MetaLinks to make sure they survive image restart). >> >> The last thing that it provides is that we do have a quite powerful >> mapping between bytecode/text/context >> and the AST. Regardless how you navigate, you get the same object. >> >> e.g. even this one works: >> >> [ 1+2 ] sourceNode == thisContext method ast blockNodes first >> >> > For example, >> > when, if ever, would a user want to access a CM's #ast (as opposed to >> > #parseTree) and could modifying it create problems? >> > >> >> Modification is a problem, yes.. code that wants to modify the AST >> without making sure the compiledMethod is in sync later >> should use #parseTree. >> >> Code that does not modify the AST (or makes sure to compile it after >> modification) is free to use #ast. >> or if you want to annotate the AST (which is a modification, after all). >> >> This is not perfect (not at all…) but the simplest solution to get (to >> some extend) what you would have if the system would have >> a real persistent, first class AST… >> >> To be improved. The ASTCache with it’s naive “lets just cache everything >> till the next image save” was done with the idea to see >> when it would show that it is too naive… for that it worked amazingly >> well till now. >> >> Marcus >> > >
