GSoC 2022: Manas Sivakumar Third and Fourth Week

[Manas23601]

Changes proposed in this pull request:

• Updated Ubuntu Github workflow to build OR-Tools in C++.
• Updated knapsack documentation.
• Temporarily not building OR-Tools with CMake.
• Ensuring Code-Flow in Knapsack Driver files.
• Implemented knapsack struct to hold data gathered from the SQL query execution.

[Manas23601]

for knapsack_driver.h in include/or-tools do I have to make my own c_type structures as well?
I have a few question regarding the implementation:

1. the knapsack function takes 3 arguments:

• bin capacity
• weights
• values
  For now I have set the weights and values to be SQL text so that users can fetch data from tables. I believe no such tables related to this (have weights and value fields) already exist. I think we don't any structures that have access to other data points like coordinates, vehichle_id, etc or do we (any mandatory arguments )? . Now this raises two questions to me. 1) Do I really need my own structs and 2) Should I change the SQL data inputs such that weights and values are arrays of integers with 'N' items.

[krashish8]

Q: Why do you want to access other data points like coordinates, vehicle_id for your knapsack function?

For the structure of your function, there are following possibilities:

• Both weights and values are arrays of integers.
• A table has 2 columns [weight, value] and the data is fetched from an SQL query.
• You have 2 separate tables for weight and value, and you run 2 separate SQL queries to fetch the data.

I think the third approach is not a good one, because why run an SQL query for a table having a single column?

You can go with either the 1st or the 2nd option. The 1st approach is intuitive because say a user has a table with all the item details, then for the knapsack function, the user can fetch two columns from that table. The 2nd approach is also okay, because any way you just require weight and value for the knapsack, and don't need any additional data, so both of these values can be fed in the form of an array. Though in this approach, you need to do an extra check that the length of weight and value arrays are the same.

Yes, if such a struct does not exist, then you have to create it.

[cvvergara]

Putting this kind of comments that are based on the sql, are useless in the C/C++ code
becuase if you change the sql then you need to look for the comments and modify accordingly.
it just implies more work for you

[Manas23601]

Q: Why do you want to access other data points like coordinates, vehicle_id for your knapsack function?

For the structure of your function, there are following possibilities:

• Both weights and values are arrays of integers.
• A table has 2 columns [weight, value] and the data is fetched from an SQL query.
• You have 2 separate tables for weight and value, and you run 2 separate SQL queries to fetch the data.

I think the third approach is not a good one, because why run an SQL query for a table having a single column?

You can go with either the 1st or the 2nd option. The 1st approach is intuitive because say a user has a table with all the item details, then for the knapsack function, the user can fetch two columns from that table. The 2nd approach is also okay, because any way you just require weight and value for the knapsack, and don't need any additional data, so both of these values can be fed in the form of an array. Though in this approach, you need to do an extra check that the length of weight and value arrays are the same.

Yes, if such a struct does not exist, then you have to create it.

My question was not very clear. I updated it now. But I think you answered my question. Is there any need for a table with additional data columns other than weights and values ? I assume if people want to use the or-tools knapsack implementation they will be having a huge number of items so asking them to enter it as parameters would be difficult