This adds a few extra settings to the default .helmignore file. In
doing this, I found a bug that some directory patterns are not
evaluated correctly. Fixed that and added tests.
Closes#989Closes#1027
This changes "pkg/chartutil".Files.Get to return a string, removes
"pkg/chartutil".Files.GetString, and adds
"pkg/chartutil".Files.GetBytes.
Closes#1020
This fixes a bug in which passed-in values files were not correctly
merged into the chart's default values YAML data. I believe it also
fixes some other prioritization bugs in values merging.
The existing unit test was wrong (see TestCoalesceValues). It is
fixed now. Also added more tests to simulate issue #971.
In the course of writing this, I removed some vestigial code as
mentioned in #920.
Closes#971Closes#920
Handle a previously unhandled error in the linter. This simply bails out
if a chart's values files do not parse.
Also, changed the implementation of CoalesceValues to return a map even
on error.
This allows templates to access information about the template file.
Right now, the template can only access the .Template.Name, which is the
chart-relative path to the current template.
Closes#894
This makes the Table() method more flexible than the original version.
It allows either a map[string]interface{} or a chartutil.Values to be
treated as a table.
Template rules
Adding chart errors
Added function that checks the existence of all the values in the templates
Adding chartfile unit tests
Testing runLinterRule
Fixing out of range
Fixing out of range
Improving quote detector
This adds support for .helmignore files. These files roughly follow
the conventions established for .gitignore files:
https://git-scm.com/docs/gitignoreCloses#748
This provides support for "global" variables. It does this by
declaring "global" to be a special namespace. It then copies this
namespace into every subchart, coalescing it into any "global"
namespace found there.
The net result is that if "global.foo" is set in the YAML file, it
will be available to every chart/subchart as ".global.foo" regardless of
where that chart is in the subchart tree.
This is a refactor of the loader to use in-memory buffers instead of
trying to optimize for memory usage by delaying reads until the last
possible moment. Since charts tend to stay well below 1M in size, this
makes more sense, and is easier to work with.