Simon, I have deleted my old implementation of modifyField! and have now switched over to your implementation, which is really nice. Please view the latest commit:
https://github.com/StephenVavasis/Modifyfield.jl It is not quite working yet, and perhaps you can help. The issue is that the macros @modify_field! and @modify_tuple_entry! are not being exported by the package even though they are declared for export. Here is the error message. julia> include(Pkg.dir("Modifyfield") * "/test_modifyfield.jl") ERROR: LoadError: UndefVarError: @modify_field! not defined in include at boot.jl:252 in include_from_node1 at loading.jl:133 while loading c:\Users\vavasis\Documents\.julia\v0.4\Modifyfield\test_modifyfiel d.jl, in expression starting on line 11 Oddly, the exporting works fine when I load the package manually (i.e., include("c:/.../Modifyfield.jl")) instead of automatically (i.e., Pkg.clone("https://github.com/StephenVavasis./Modifyfield.jl";)) . I have never written a package with macros before, so perhaps I am making some foolish mistake? Beyond this, I have a few comments/questions: (1) Although I didn't use it, regarding your suggestion of "eltype" to determine the base type of an array: Unfortunately, this does not work in general because eltype in general returns the type that comes out of a start/done/next loop rather than a getindex call. For example, eltype(Dict{K,V}()) == (K,V) (2) I did not understand the use of symbol in :call, symbol(x) in the code that you provided. In fact, :call, x works fine because in Julia you can call a type to get the constructor. Am I missing something? (3) I also did not understand why you put esc around argument x but not v in your code snippet. Based on my (possibly incorrect) understanding of esc, it seems like both x and v should be escaped, which is what I have implemented. Perhaps you can clarify this. (4) Finally, building upon your ingenious code, I also added a @modify_tuple_entry! macro to the package. I'd appreciate your comments on that as well. On Tuesday, June 2, 2015 at 8:16:29 AM UTC-4, Simon Byrne wrote: > > On 2 June 2015 at 13:00, David Gold <[email protected] <javascript:>> > wrote: > > @Simon: why safer? Rather, what is unsafe about calling symbol within > the Expr constructor? > > I wasn't quite sure how the scoping worked (i.e. what would happen if > the type had not been exported?), but it does seem to be safe. >
