Previously, the compiler performed an incremental build by analyzing and
resolving all classes in the program (even unchanged ones) and then using
the dependency graph information to determine which .js files were stale and
needed to be re-emitted. This algorithm produced "correct" rebuilds, but the
cost of re-analyzing the entire program turned out to be higher than
anticipated, especially for component-heavy compilations.
To achieve performant rebuilds, it is necessary to reuse previous analysis
results if possible. Doing this safely requires knowing when prior work is
viable and when it is stale and needs to be re-done.
The new algorithm implemented by this commit is such:
1) Each incremental build starts with knowledge of the last known good
dependency graph and analysis results from the last successful build,
plus of course information about the set of files changed.
2) The previous dependency graph's information is used to determine the
set of source files which have "logically" changed. A source file is
considered logically changed if it or any of its dependencies have
physically changed (on disk) since the last successful compilation. Any
logically unchanged dependencies have their dependency information copied
over to the new dependency graph.
3) During the `TraitCompiler`'s loop to consider all source files in the
program, if a source file is logically unchanged then its previous
analyses are "adopted" (and their 'register' steps are run). If the file
is logically changed, then it is re-analyzed as usual.
4) Then, incremental build proceeds as before, with the new dependency graph
being used to determine the set of files which require re-emitting.
This analysis reuse avoids template parsing operations in many circumstances
and significantly reduces the time it takes ngtsc to rebuild a large
application.
Future work will increase performance even more, by tackling a variety of
other opportunities to reuse or avoid work.
PR Close#34288
Previously, our incremental build system kept track of the changes between
the current compilation and the previous one, and used its knowledge of
inter-file dependencies to evaluate the impact of each change and emit the
right set of output files.
However, a problem arose if the compiler was not able to extract a
dependency graph successfully. This typically happens if the input program
contains errors. In this case the Angular analysis part of compilation is
never executed.
If a file changed in one of these failed builds, in the next build it
appears unchanged. This means that the compiler "forgets" to emit it!
To fix this problem, the compiler needs to know the set of changes made
_since the last successful build_, not simply since the last invocation.
This commit changes the incremental state system to much more explicitly
pass information from the previous to the next compilation, and in the
process to keep track of changes across multiple failed builds, until the
program can be analyzed successfully and the results of those changes
incorporated into the emit plan.
Fixes#32214
PR Close#33971
Previously, the ngtsc compiler attempted to reuse analysis work from the
previous program during an incremental build. To do this, it had to prove
that the work was safe to reuse - that no changes made to the new program
would invalidate the previous analysis.
The implementation of this had a significant design flaw: if the previous
program had errors, the previous analysis would be missing significant
information, and the dependency graph extracted from it would not be
sufficient to determine which files should be re-analyzed to fill in the
gaps. This often meant that the build output after an error was resolved
would be wholly incorrect.
This commit switches ngtsc to take a simpler approach to incremental
rebuilds. Instead of attempting to reuse prior analysis work, the entire
program is re-analyzed with each compilation. This is actually not as
expensive as one might imagine - analysis is a fairly small part of overall
compilation time.
Based on the dependency graph extracted during this analysis, the compiler
then can make accurate decisions on whether to emit specific files. A new
suite of tests is added to validate behavior in the presence of source code
level errors.
This new approach is dramatically simpler than the previous algorithm, and
should always produce correct results for a semantically correct program.s
Fixes#32388Fixes#32214
PR Close#33862
Joey Perrott
Alex Rickabaugh