diff --git a/docs/UnityAssertionsReference.md b/docs/UnityAssertionsReference.md index 99880d8..0a0e51b 100644 --- a/docs/UnityAssertionsReference.md +++ b/docs/UnityAssertionsReference.md @@ -10,46 +10,46 @@ Upon boolean False, an assertion stops execution and reports the failure. and easily execute those assertions. - The structure of Unity allows you to easily separate test assertions from source code in, well, test code. -- Unity's assertions: -- Come in many, many flavors to handle different C types and assertion cases. -- Use context to provide detailed and helpful failure messages. -- Document types, expected values, and basic behavior in your source code for +- Unity’s assertions: + - Come in many, many flavors to handle different C types and assertion cases. + - Use context to provide detailed and helpful failure messages. + - Document types, expected values, and basic behavior in your source code for free. -### Unity Is Several Things But Mainly It's Assertions +### Unity Is Several Things But Mainly It’s Assertions One way to think of Unity is simply as a rich collection of assertions you can use to establish whether your source code behaves the way you think it does. Unity provides a framework to easily organize and execute those assertions in test code separate from your source code. -### What's an Assertion? +### What’s an Assertion? At their core, assertions are an establishment of truth - boolean truth. Was this thing equal to that thing? Does that code doohickey have such-and-such property -or not? You get the idea. Assertions are executable code (to appreciate the big -picture on this read up on the difference between -[link:Dynamic Verification and Static Analysis]). A failing assertion stops -execution and reports an error through some appropriate I/O channel (e.g. -stdout, GUI, file, blinky light). +or not? You get the idea. Assertions are executable code. Static analysis is a +valuable approach to improving code quality, but it is not executing your code +in the way an assertion can. A failing assertion stops execution and reports an +error through some appropriate I/O channel (e.g. stdout, GUI, output file, +blinky light). Fundamentally, for dynamic verification all you need is a single assertion -mechanism. In fact, that's what the [assert() macro][] in C's standard library +mechanism. In fact, that’s what the [assert() macro][] in C’s standard library is for. So why not just use it? Well, we can do far better in the reporting -department. C's `assert()` is pretty dumb as-is and is particularly poor for +department. C’s `assert()` is pretty dumb as-is and is particularly poor for handling common data types like arrays, structs, etc. And, without some other -support, it's far too tempting to litter source code with C's `assert()`'s. It's +support, it’s far too tempting to litter source code with C’s `assert()`’s. It’s generally much cleaner, manageable, and more useful to separate test and source code in the way Unity facilitates. -### Unity's Assertions: Helpful Messages _and_ Free Source Code Documentation +### Unity’s Assertions: Helpful Messages _and_ Free Source Code Documentation Asserting a simple truth condition is valuable, but using the context of the -assertion is even more valuable. For instance, if you know you're comparing bit +assertion is even more valuable. For instance, if you know you’re comparing bit flags and not just integers, then why not use that context to give explicit, readable, bit-level feedback when an assertion fails? -That's what Unity's collection of assertions do - capture context to give you +That’s what Unity’s collection of assertions do - capture context to give you helpful, meaningful assertion failure messages. In fact, the assertions themselves also serve as executable documentation about types and values in your source code. So long as your tests remain current with your source and all those @@ -73,12 +73,12 @@ a simple null check). - `Actual` is the value being tested and unlike the other parameters in an assertion construction is the only parameter present in all assertion variants. - `Modifiers` are masks, ranges, bit flag specifiers, floating point deltas. -- `Expected` is your expected value (duh) to compare to an `actual` value; it's +- `Expected` is your expected value (duh) to compare to an `actual` value; it’s marked as an optional parameter because some assertions only need a single `actual` parameter (e.g. null check). - `Size/count` refers to string lengths, number of array elements, etc. -Many of Unity's assertions are clear duplications in that the same data type +Many of Unity’s assertions are clear duplications in that the same data type is handled by several assertions. The differences among these are in how failure messages are presented. For instance, a `_HEX` variant of an assertion prints the expected and actual values of that assertion formatted as hexadecimal. @@ -99,7 +99,7 @@ _Example:_ TEST_ASSERT_X( {modifiers}, {expected}, actual, {size/count} ) ``` -becomes messageified like thus... +becomes messageified like thus… ```c TEST_ASSERT_X_MESSAGE( {modifiers}, {expected}, actual, {size/count}, message ) @@ -108,7 +108,7 @@ TEST_ASSERT_X_MESSAGE( {modifiers}, {expected}, actual, {size/count}, message ) Notes: - The `_MESSAGE` variants intentionally do not support `printf` style formatting - since many embedded projects don't support or avoid `printf` for various reasons. + since many embedded projects don’t support or avoid `printf` for various reasons. It is possible to use `sprintf` before the assertion to assemble a complex fail message, if necessary. - If you want to output a counter value within an assertion fail message (e.g. from @@ -119,7 +119,7 @@ Notes: Unity provides a collection of assertions for arrays containing a variety of types. These are documented in the Array section below. These are almost on par -with the `_MESSAGE`variants of Unity's Asserts in that for pretty much any Unity +with the `_MESSAGE`variants of Unity’s Asserts in that for pretty much any Unity type assertion you can tack on `_ARRAY` and run assertions on an entire block of memory. @@ -144,7 +144,7 @@ Notes: Unity provides a collection of assertions for arrays containing a variety of types which can be compared to a single value as well. These are documented in the Each Equal section below. these are almost on par with the `_MESSAGE` -variants of Unity's Asserts in that for pretty much any Unity type assertion you +variants of Unity’s Asserts in that for pretty much any Unity type assertion you can inject `_EACH_EQUAL` and run assertions on an entire block of memory. ```c @@ -203,7 +203,7 @@ code then verifies as a final step. #### `TEST_PASS_MESSAGE("message")` This will abort the remainder of the test, but count the test as a pass. Under -normal circumstances, it is not necessary to include this macro in your tests... +normal circumstances, it is not necessary to include this macro in your tests… a lack of failure will automatically be counted as a `PASS`. It is occasionally useful for tests with `#ifdef`s and such. @@ -392,7 +392,7 @@ Asserts that the pointers point to the same memory location. #### `TEST_ASSERT_EQUAL_STRING (expected, actual)` -Asserts that the null terminated (`'\0'`)strings are identical. If strings are +Asserts that the null terminated (`’\0’`)strings are identical. If strings are of different lengths or any portion of the strings before their terminators differ, the assertion fails. Two NULL strings (i.e. zero length) are considered equivalent. @@ -561,7 +561,7 @@ Asserts that the `actual` value is NOT within +/- `delta` of the `expected` valu #### `TEST_ASSERT_EQUAL_FLOAT (expected, actual)` -Asserts that the `actual` value is "close enough to be considered equal" to the +Asserts that the `actual` value is “close enough to be considered equal” to the `expected` value. If you are curious about the details, refer to the Advanced Asserting section for more details on this. Omitting a user-specified delta in a floating point assertion is both a shorthand convenience and a requirement of @@ -569,7 +569,7 @@ code generation conventions for CMock. #### `TEST_ASSERT_NOT_EQUAL_FLOAT (expected, actual)` -Asserts that the `actual` value is NOT "close enough to be considered equal" to the +Asserts that the `actual` value is NOT “close enough to be considered equal” to the `expected` value. #### `TEST_ASSERT_FLOAT_ARRAY_WITHIN (delta, expected, actual, num_elements)` @@ -662,7 +662,7 @@ Asserts that the `actual` value is NOT within +/- `delta` of the `expected` valu #### `TEST_ASSERT_EQUAL_DOUBLE (expected, actual)` -Asserts that the `actual` value is "close enough to be considered equal" to the +Asserts that the `actual` value is “close enough to be considered equal” to the `expected` value. If you are curious about the details, refer to the Advanced Asserting section for more details. Omitting a user-specified delta in a floating point assertion is both a shorthand convenience and a requirement of @@ -670,7 +670,7 @@ code generation conventions for CMock. #### `TEST_ASSERT_NOT_EQUAL_DOUBLE (expected, actual)` -Asserts that the `actual` value is NOT "close enough to be considered equal" to the +Asserts that the `actual` value is NOT “close enough to be considered equal” to the `expected` value. #### `TEST_ASSERT_DOUBLE_ARRAY_WITHIN (delta, expected, actual, num_elements)` @@ -753,7 +753,7 @@ Not A Number floating point representations. This section helps you understand how to deal with some of the trickier assertion situations you may run into. It will give you a glimpse into some of -the under-the-hood details of Unity's assertion mechanisms. If you're one of +the under-the-hood details of Unity’s assertion mechanisms. If you’re one of those people who likes to know what is going on in the background, read on. If not, feel free to ignore the rest of this document until you need it. @@ -768,9 +768,9 @@ mathematical operations might result in a representation of 8 x 2-2 that also evaluates to a value of 2. At some point repeated operations cause equality checks to fail. -So Unity doesn't do direct floating point comparisons for equality. Instead, it -checks if two floating point values are "really close." If you leave Unity -running with defaults, "really close" means "within a significant bit or two." +So Unity doesn’t do direct floating point comparisons for equality. Instead, it +checks if two floating point values are “really close.” If you leave Unity +running with defaults, “really close” means “within a significant bit or two.” Under the hood, `TEST_ASSERT_EQUAL_FLOAT` is really `TEST_ASSERT_FLOAT_WITHIN` with the `delta` parameter calculated on the fly. For single precision, delta is the expected value multiplied by 0.00001, producing a very small proportional @@ -779,28 +779,27 @@ range around the expected value. If you are expecting a value of 20,000.0 the delta is calculated to be 0.2. So any value between 19,999.8 and 20,000.2 will satisfy the equality check. This works out to be roughly a single bit of range for a single-precision number, and -that's just about as tight a tolerance as you can reasonably get from a floating +that’s just about as tight a tolerance as you can reasonably get from a floating point value. -So what happens when it's zero? Zero - even more than other floating point -values - can be represented many different ways. It doesn't matter if you have -0 x 20 or 0 x 263.It's still zero, right? Luckily, if you -subtract these values from each other, they will always produce a difference of -zero, which will still fall between 0 plus or minus a delta of 0. So it still -works! +So what happens when it’s zero? Zero - even more than other floating point +values - can be represented many different ways. It doesn’t matter if you have +0x20 or 0x263. It’s still zero, right? Luckily, if you subtract these +values from each other, they will always produce a difference of zero, which +will still fall between 0 plus or minus a delta of 0. So it still works! Double precision floating point numbers use a much smaller multiplier, again approximating a single bit of error. -If you don't like these ranges and you want to make your floating point equality +If you don’t like these ranges and you want to make your floating point equality assertions less strict, you can change these multipliers to whatever you like by defining UNITY_FLOAT_PRECISION and UNITY_DOUBLE_PRECISION. See Unity documentation for more. ### How do we deal with targets with non-standard int sizes? -It's "fun" that C is a standard where something as fundamental as an integer -varies by target. According to the C standard, an `int` is to be the target's +It’s “fun” that C is a standard where something as fundamental as an integer +varies by target. According to the C standard, an `int` is to be the target’s natural register size, and it should be at least 16-bits and a multiple of a byte. It also guarantees an order of sizes: @@ -814,7 +813,7 @@ and this remains perfectly standard C. To make things even more interesting, there are compilers and targets out there that have a hard choice to make. What if their natural register size is 10-bits -or 12-bits? Clearly they can't fulfill _both_ the requirement to be at least +or 12-bits? Clearly they can’t fulfill _both_ the requirement to be at least 16-bits AND the requirement to match the natural register size. In these situations, they often choose the natural register size, leaving us with something like this: @@ -823,24 +822,24 @@ something like this: char (8 bit) <= short (12 bit) <= int (12 bit) <= long (16 bit) ``` -Um... yikes. It's obviously breaking a rule or two... but they had to break SOME +Um… yikes. It’s obviously breaking a rule or two… but they had to break SOME rules, so they made a choice. When the C99 standard rolled around, it introduced alternate standard-size types. It also introduced macros for pulling in MIN/MAX values for your integer types. -It's glorious! Unfortunately, many embedded compilers can't be relied upon to +It’s glorious! Unfortunately, many embedded compilers can’t be relied upon to use the C99 types (Sometimes because they have weird register sizes as described -above. Sometimes because they don't feel like it?). +above. Sometimes because they don’t feel like it?). A goal of Unity from the beginning was to support every combination of -microcontroller or microprocessor and C compiler. Over time, we've gotten really +microcontroller or microprocessor and C compiler. Over time, we’ve gotten really close to this. There are a few tricks that you should be aware of, though, if -you're going to do this effectively on some of these more idiosyncratic targets. +you’re going to do this effectively on some of these more idiosyncratic targets. -First, when setting up Unity for a new target, you're going to want to pay +First, when setting up Unity for a new target, you’re going to want to pay special attention to the macros for automatically detecting types (where available) or manually configuring them yourself. You can get information -on both of these in Unity's documentation. +on both of these in Unity’s documentation. What about the times where you suddenly need to deal with something odd, like a 24-bit `int`? The simplest solution is to use the next size up. If you have a @@ -848,9 +847,9 @@ What about the times where you suddenly need to deal with something odd, like a `int`, configure Unity to use 16 bits. There are two ways this is going to affect you: -1. When Unity displays errors for you, it's going to pad the upper unused bits +1. When Unity displays errors for you, it’s going to pad the upper unused bits with zeros. -2. You're going to have to be careful of assertions that perform signed +2. You’re going to have to be careful of assertions that perform signed operations, particularly `TEST_ASSERT_INT_WITHIN`. Such assertions might wrap your `int` in the wrong place, and you could experience false failures. You can always back down to a simple `TEST_ASSERT` and do the operations yourself.