add more instructions for generateprecompiles.jl

This commit is contained in:
Leon Shen 2019-11-14 14:00:39 -05:00
parent 05eed0fdda
commit 70c507b8f8

View File

@ -1,6 +1,29 @@
# To figure out what should be precompiled, run this script, then move
# precompile_Plots.jl in precompiles_path (see below) to src/precompile.jl
# This script works by using SnoopCompile to log compilations that take place
# while running the examples on the GR backend. So SnoopCompile must be
# installed, and StatsPlots, RDatasets, and FileIO are also required for
# certain examples.
# If precompilation fails with an UndefVarError for a module, probably what is
# happening is that the module appears in the precompile statements, but is
# only visible to one of Plots' dependencies, and not Plots itself. Adding the
# module to the blacklist below will remove these precompile statements.
# Anonymous functions may appear in precompile statements as functions with
# hashes in their name. Those of the form "#something##kw" have to do with
# compiling functions with keyword arguments, and are named reproducibly, so
# can be kept. Others generally will not work. Currently, SnoopCompile includes
# some anonymous functions that not reproducible, but SnoopCompile PR #30
# (which looks about to be merged) will ensure that anonymous functions are
# actually defined before attempting to precompile them. Alternatively, we can
# keep only the keyword argument related anonymous functions by changing the
# regex that SnoopCompile uses to detect anonymous functions to
# r"#{1,2}[^\"#]+#{1,2}\d+" (see anonrex in SnoopCompile.jl). To exclude all
# precompile statements involving anonymous functions, "#" can also be added to
# the blacklist below.
using SnoopCompile
log_path = joinpath(tempdir(), "compiles.log")