Create fields in hook_install.

We have tests now, changing to Needs Review to get testbots opinion on the patch.

+++ b/commerce_discount.install
@@ -126,19 +133,17 @@ function commerce_discount_requirements($phase) {
+  field_info_cache_clear();
...
-    field_info_cache_clear();
+  if (!isset($field_types['entityreference'], $field_types['inline_conditions'], $field_types['commerce_product_reference'])) {

This was a performance improvement for clearing cache because install_helper is called when clearing cache too.

Not against this patch, it's has some promise! thanks for posting @gaydamaka

Used it on my a project - works really good! Let's have it merged.

1. +++ b/commerce_discount.install
   @@ -6,6 +6,14 @@
   +function commerce_discount_install() {
   +  field_info_cache_clear();
   +  commerce_discount_install_helper();
   +}
   

   All good with committing this without the field_info_cache_clear() call.

2. +++ b/commerce_discount.install
   @@ -129,15 +137,6 @@ function commerce_discount_install_helper() {
   -  if (!isset($field_types['entityreference'])
   -    || !isset($field_types['inline_conditions'])
   -    || !isset($field_types['commerce_product_reference'])
   -  ) {
   -    field_info_cache_clear();
   -  }
   

   Question: Is there something wrong with this logic for it to be removed? Again this also is called on cache flushes so it likely needs to stay.

Want to see it merged.

Would have been nice if you two addressed my questions in #9 instead of throwing patches and RTBCing them. If just to instill a bit more confidence behind the patch.

Though I've thoroughly reviewed all the code paths in the changes and there should be no unforeseen problems with the latest.

Was a bit hesitant on the multiple isset() call change there at first too, because it's uncommon in core and could be seen as hard to read unless you know that isset() with multiple arguments returns FALSE on the first FALSE value and is short circuited the same as the || operator would.

Committed and pushed to latest dev branch.