fix: always do type check for all transformed files

[ahnpnl]

closes #1390

Previously we always do type checking while compiling codes. However, when a file is cached and the next time when running tests, our compiling logic won't be invoked when the file content (which is provided by jest as input for getCacheKey) is the same as cached content therefore the unchanged file won't be type checking.

So the change is moving the part of type checking to the entry point getCacheKey which jest uses to determine whether a file should be recompiled or simply get cached content.

[coveralls]

Pull Request Test Coverage Report for Build 4189

• 28 of 28 (100.0%) changed or added relevant lines in 5 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.01%) to 92.631%

---

Totals
Change from base Build 4188:	-0.01%
Covered Lines:	1123
Relevant Lines:	1160

---

💛 - Coveralls

[kirillgroshkov]

Will it still respect the diagnostics: false config option?

For me, after updating to this version, I see the increased memory heap consumption (as if it was not using jest cache). Basically, with ts-jest@25.2.1 with warm cache my first tests reported ~100Mb heap (or ~500Mb with cold cache). ts-jest@25.3.0 report ~500Mb on first test files regardless of warm or cold cache.

[ahnpnl]

yes, this PR still respect diagnostics: false.

[kirillgroshkov]

yes, this PR still respect diagnostics: false.

Got it. Then my observations are probably caused by something else. Will keep an eye on the new version and report if I find anything else suspicious.

Thanks for the great work BTW!

[ahnpnl]

watch mode fix will have to wait :) this PR only fixed the issue for non watch mode

[ahnpnl]

@kirillgroshkov besides the memory part, how do you see the speed of no cache in 25.3.0 comparing to 25.2.1 ?

[kirillgroshkov]

The speed, in terms of how many seconds it takes to complete all tests is more or less identical, so, no change between 25.2.1 and 25.3.0

[ahnpnl]

thanks :) i'm glad to hear that.