Improve substr() performance by avoiding using owned string

[richox]

Which issue does this PR close?

Closes #13687 .

Rationale for this change

What changes are included in this PR?

Are these changes tested?

tested with: cargo bench --bench substr
the result shows that substr_string and substr_large_string are running much faster in all cases.

   Compiling datafusion-functions v43.0.0 (/Volumes/Workspace/blaze-init/arrow-datafusion/datafusion/functions)
    Finished `bench` profile [optimized] target(s) in 2m 15s
     Running benches/substr.rs (target/release/deps/substr-a86cfe081ca878bf)
Gnuplot not found, using plotters backend
SHORTER THAN 12/substr_string_view [size=1024, strlen=12]
                        time:   [17.906 µs 18.033 µs 18.166 µs]
                        change: [-5.6189% -4.0418% -2.4317%] (p = 0.00 < 0.05)
                        Performance has improved.
SHORTER THAN 12/substr_string [size=1024, strlen=12]
                        time:   [16.450 µs 16.554 µs 16.670 µs]
                        change: [-71.933% -71.626% -71.291%] (p = 0.00 < 0.05)
                        Performance has improved.
SHORTER THAN 12/substr_large_string [size=1024, strlen=12]
                        time:   [18.539 µs 18.803 µs 19.144 µs]
                        change: [-70.591% -69.829% -69.005%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high severe

LONGER THAN 12/substr_string_view [size=1024, count=64, strlen=128]
                        time:   [28.407 µs 28.605 µs 28.818 µs]
                        change: [-2.8033% -0.9698% +0.5511%] (p = 0.33 > 0.05)
                        No change in performance detected.
LONGER THAN 12/substr_string [size=1024, count=64, strlen=128]
                        time:   [29.680 µs 29.874 µs 30.084 µs]
                        change: [-64.029% -63.622% -63.202%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 10 measurements (30.00%)
  2 (20.00%) low mild
  1 (10.00%) high mild
LONGER THAN 12/substr_large_string [size=1024, count=64, strlen=128]
                        time:   [29.907 µs 30.260 µs 30.683 µs]
                        change: [-62.955% -61.999% -61.035%] (p = 0.00 < 0.05)
                        Performance has improved.

SRC_LEN > 12, SUB_LEN < 12/substr_string_view [size=1024, count=6, strlen=128]
                        time:   [26.295 µs 26.483 µs 26.682 µs]
                        change: [-5.6813% -4.7257% -3.7561%] (p = 0.00 < 0.05)
                        Performance has improved.
SRC_LEN > 12, SUB_LEN < 12/substr_string [size=1024, count=6, strlen=128]
                        time:   [29.560 µs 29.838 µs 30.140 µs]
                        change: [-59.939% -59.336% -58.767%] (p = 0.00 < 0.05)
                        Performance has improved.
SRC_LEN > 12, SUB_LEN < 12/substr_large_string [size=1024, count=6, strlen=128]
                        time:   [30.994 µs 31.199 µs 31.413 µs]
                        change: [-57.768% -57.255% -56.796%] (p = 0.00 < 0.05)
                        Performance has improved.

SHORTER THAN 12/substr_string_view [size=4096, strlen=12]
                        time:   [71.298 µs 71.768 µs 72.296 µs]
                        change: [-3.9430% -2.9409% -1.9043%] (p = 0.00 < 0.05)
                        Performance has improved.
SHORTER THAN 12/substr_string [size=4096, strlen=12]
                        time:   [64.237 µs 64.725 µs 65.242 µs]
                        change: [-72.971% -71.944% -71.285%] (p = 0.00 < 0.05)
                        Performance has improved.
SHORTER THAN 12/substr_large_string [size=4096, strlen=12]
                        time:   [71.706 µs 72.637 µs 74.080 µs]
                        change: [-70.310% -69.729% -69.034%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high severe

LONGER THAN 12/substr_string_view [size=4096, count=64, strlen=128]
                        time:   [111.94 µs 113.43 µs 115.24 µs]
                        change: [-2.1479% -0.4786% +1.4386%] (p = 0.64 > 0.05)
                        No change in performance detected.
Found 2 outliers among 10 measurements (20.00%)
  1 (10.00%) high mild
  1 (10.00%) high severe
LONGER THAN 12/substr_string [size=4096, count=64, strlen=128]
                        time:   [114.55 µs 116.05 µs 118.04 µs]
                        change: [-64.142% -63.506% -62.753%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high mild
LONGER THAN 12/substr_large_string [size=4096, count=64, strlen=128]
                        time:   [116.93 µs 117.71 µs 118.67 µs]
                        change: [-62.124% -61.458% -60.871%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high severe

SRC_LEN > 12, SUB_LEN < 12/substr_string_view [size=4096, count=6, strlen=128]
                        time:   [104.15 µs 105.34 µs 106.74 µs]
                        change: [-5.7933% -4.3114% -2.7804%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high mild
SRC_LEN > 12, SUB_LEN < 12/substr_string [size=4096, count=6, strlen=128]
                        time:   [114.42 µs 115.86 µs 117.83 µs]
                        change: [-60.701% -59.738% -58.852%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high severe
SRC_LEN > 12, SUB_LEN < 12/substr_large_string [size=4096, count=6, strlen=128]
                        time:   [122.53 µs 123.66 µs 125.01 µs]
                        change: [-56.827% -56.179% -55.534%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 10 measurements (20.00%)
  2 (20.00%) high mild

Are there any user-facing changes?

no.

[jonahgao]

CI failure should be caused by #13686

[Dandandan]

I wonder if we can see perf change on TPC-H query 22 (which includes some substring expressions).

[richox]

CI failure should be caused by #13686

it seems cargo test still fails on unrelated cases.

[jonahgao]

it seems cargo test still fails on unrelated cases.

I think merging the main branch can fix it.

[jonahgao]

LGTM, thanks @richox .

TPC-H query 22 uses string_view_substr, seems not affected by this PR. cc @Dandandan

[Dandandan]

LGTM, thanks @richox .

TPC-H query 22 uses string_view_substr, seems not affected by this PR. cc @Dandandan

Ah makes sense

[alamb]

Thanks @richox @jonahgao and @Dandandan