[DRAFT] Vector Data-Structure #279
Conversation
|
Please note that the main development will first happen in the fork https://revol-xut/rtrlib, and then we will do a larger PR that adds full ASPA support at once, following the rfc @waehlisch just posted. So this PR will grow slowly over the next couple of weeks. |
cadc145 to
7f1ea85
Compare
|
Several quick comments:
I have not really looked at the code so far, I will do that after you have explained the why. |
7f1ea85 to
9ac1107
Compare
- add aspa_record struct
- add aspa_vector_{create, delete, reallocate, ordered_insert, insert, delete_at, find_cas}
- aspa_vector is simimlar in function to std::vector but is sorted therefore search can be performed with binary search
9ac1107 to
a57f9bd
Compare
There was a problem hiding this comment.
great project, the code looks good, I reviewed ~80.
- I'm wondering why it is chosen to dynamically resize the array.
Why does it fit better the user case then then letting the caller specify the
initial size + providing a resize operation where the new size is explicitly
passed?
The reason should also be documented in the commit and/or PR description. - It's not needed to have a comment for every line.
Most of the the code is self-explanatory, e.g. when callingfreeor assigning an array element. Having comments for those simple operations makes it actually more cumbersome to read. - The naming became inconsistent. In some places the data structure is called
vector (var names, documentations) in other places ordered_dyn_array. - Use more error codes then -1.
Currently only -1 is returned on errors. This makes it very hard to impossible for the caller to figure why an operation failed.
| int ordered_dyn_array_create(struct ordered_dyn_array **vector_pointer) | ||
| { | ||
| const size_t DEFAULT_INITIAL_SIZE = 128; | ||
|
|
There was a problem hiding this comment.
If you allocate *vector first you do not need the data_field var and could directory store the pointer for the allocated data memory in vector->data
|
|
||
| int ordered_dyn_array_create(struct ordered_dyn_array **vector_pointer) | ||
| { | ||
| const size_t DEFAULT_INITIAL_SIZE = 128; |
| if (vector->data != NULL) { | ||
| // freeing the data | ||
| free(vector->data); | ||
| } |
There was a problem hiding this comment.
Why is the behavior different when vector is NULL and vector->data is NULL?
If vector is null an error is returned, if vector->data is null it will be silently ignored and success will be returned.
Is there a valid usecase where vector is not null but vector->data is?
| struct aspas_record *new_data_field = (struct aspas_record *)lrtr_malloc( | ||
| sizeof(struct aspas_record) * vector->capacity * SIZE_INCREASE_EXPONENTIAL); | ||
|
|
||
| // malloc failed so returning an error | ||
| if (new_data_field == NULL) { | ||
| return -1; | ||
| } | ||
|
|
||
| // copying the data from the old location to the new one | ||
| memcpy(new_data_field, vector->data, vector->capacity * sizeof(struct aspas_record)); | ||
|
|
||
| // deleting the old vector | ||
| free(vector->data); | ||
|
|
||
| // assigning the new array to the vector and incrementing the capacity | ||
| vector->data = new_data_field; |
There was a problem hiding this comment.
This block looks like it could be replaced by a single call to realloc or reallocarray.
| * creates an vector object | ||
| * @param[vector_pointer] the pointer to the newly created pointer will be written to *vector_pointer | ||
| * @result Valid pointer to an ordered_dyn_array struct | ||
| * @result Null On error. |
There was a problem hiding this comment.
documentation is outdated, it can't return null
| * @result 0 On success. | ||
| * @result -1 On error. | ||
| */ | ||
| int ordered_dyn_array_reallocate(struct ordered_dyn_array *vector); |
There was a problem hiding this comment.
Why is this function exported?
If the user can't choose the new size, I don't know a reason why he would ever want to call it explicitly.
|
|
||
| /** | ||
| * @brief Struct which is similar in function to std::vector from C++. | ||
| * If the vector is running full a larger chunk of memory is allocated and the data is copied over. |
There was a problem hiding this comment.
document in which order the records are kept
| // if the vector is null just return | ||
| if (vector == NULL) { | ||
| return -1; | ||
| } |
There was a problem hiding this comment.
I don't think it adds value to return success/failure the nothing from this function.
If an error is returned how is the caller expected to handle the situation?
I would instead either:
- Expect that the caller calls this function only once for a vector and does not call it with a nulll value. If he does run into the segfault.
- Handle the situation gracefully if one of the ptr is null ignore it, only run free if it is not null, additionally log a message if one of the ptrs is null but it is not expected, to allow to find the cases and fix them.
| }; | ||
|
|
||
| /** | ||
| * @brief Struct which is similar in function to std::vector from C++. |
There was a problem hiding this comment.
Instead of referencing std::vector here, it would be better to describe what it does.
Otherwise the reader first have to figure out how std::vector functions to know what ordered_dyn_array does
|
@tanneberger can this PR be closed? Seems to be superseded by #285 |
I assume, no reply means "Yes". |
Contribution
Testing procedure
Issues/PRs references