ARROW-5346: [C++] Revert changed to vendored datetime library

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.

[emkornfield]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

[pravindra]

The best way I could think of accomplishing this was to introduce another top level namespace arrow_vendored, so "duration" does not conflict between the two. Other suggestions welcome.

lgtm.

[emkornfield]

Looks like I missed a gandiva file, will fix tomorrow.

[pitrou]

Have you tried reporting the issue upstream?

[wesm]

I agree that the issue should get fixed upstream. If we are having this arrow_vendored namespace (which actually seems like a good idea to isolate the vendored implementations from namespace contamination) then the util:: is probably superfluous

[emkornfield]

I can ask for a fix upstream, but it isn't small and it really is an edge case with how we wrapped the library into our own namespace. The only other way this occurs is if someone has "duration" included in a file for before "date.h" without any namespace.

Essentially the code boiled down to:

namespace b {
void print_me() {
std::cout << "buz";
}
}
namespace f {
void print_me() {
std::cout << "foo";
}
void run() {
using namespace b;
print_me();
}
}

So print_me was getting resolved to f::print_me (in our case arrow::duration) instead of b::print_me (std::chrono::duration).

[emkornfield]

Opened an issue upstream. I think this is a worthwhile fix anyways (i've removed the util namespace).

[codecov-io]

Codecov Report

Merging #4324 into master will increase coverage by 0.97%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #4324 +/- ##
=========================================
+ Coverage 88.22% 89.2% +0.97% 
=========================================
Files 777 632 -145 Lines 97881 86483 -11398 Branches 1251 0 -1251 =========================================
- Hits 86360 77147 -9213 + Misses 11285 9336 -1949 + Partials 236 0 -236

Impacted Files	Coverage Δ
cpp/src/gandiva/precompiled/time.cc	97.19% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/date.h	30.62% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz_private.h	33.33% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.h	43.47% <ø> (ø)	⬆️
cpp/src/arrow/vendored/datetime/tz.cpp	65.45% <ø> (ø)	⬆️
cpp/src/gandiva/cast_time.cc	100% <ø> (ø)	⬆️
cpp/src/gandiva/date_utils.h	100% <100%> (ø)	⬆️
cpp/src/gandiva/precompiled/epoch_time_point.h	100% <100%> (ø)	⬆️
cpp/src/arrow/util/parsing.h	96.11% <100%> (ø)	⬆️
cpp/src/arrow/pretty_print.cc	83.48% <100%> (ø)	⬆️
... and 147 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c63c79...3d2581e. Read the comment docs.