---

Pipelining unary operators σ and π

---

---

---

 

Slideshow:

Review: processing the σ operation using Table-Scan/Index-Scan
 
Input buffering requirement to use pipeling for processing σ
 
Because σ does not access the relation on disk directly, it does not need any buffers
Input buffering requirement to use pipeling for processing σ
 
Because σ does not access the relation on disk directly, it does not need any buffers
The Table-Scan operation - on the other hand - needs to use 1 buffer to store tuples read from disk.
Input buffering requirement to use pipeling for processing π
 
Because π does not access the relation on disk directly, it does not need any buffers
❮ ❯



---

---

---

---

• Prelude: σ and π do not read relations
  • Recall:

    There are 2 basic (relation) access operators: Table-Scan Index-Scan           See: click here

    Therefore:

    The implementation algorithms for σ and π do not access a relation directly (If σ/π needs to get input from a relation, then they will use a Table-Scan or Index-Scan first !!!) For example:

  ---

  ---

  ---

  ---

• Pipelining σ_c( • )
  • Pipelining a σ_c( • ) operation:

    General way to pipeline the σc( • ) operation: You only need 1 tuple (variable) as input variable because σ operates on 1 tuple at a time. Example: σC(R) Example: σ(A=c) and cond2(R) and relation R has an index on A

  ---

  ---

  ---

  ---

• Pipelining π_attrs
  • Pipelining a π_attrs( • ) operation:

    General way to pipeline the πattrs( • ) operation: You only need 1 tuple variable as "buffer" because π operates on 1 tuple at a time.