ARROW-8471: [C++][Integration] Represent 64 bit integers as strings

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352

[bkietz]

Some 64 bit integer values are not representable as JSON numbers, so store these as strings.

This is issue is a regression; original fix was #5267

[github-actions]

https://issues.apache.org/jira/browse/ARROW-8471

[fsaintjacques]

Why the change, just set GOPATH if you have something else?

[bkietz]

I don't know why ~ failed resolve to $HOME locally but this seemed like a way to save someone the next person from the frustration of reading the script to figure out that GOPATH needs set.

[kiszk]

not sure 100%. but, is this indentation correct?

[bkietz]

C++ indentation is deferred to clang-format. We have a CI job which ensures that running clang-format is a no-op. As for explaining why it made this choice... I'm not familiar enough with its implementation to say, sorry

[fsaintjacques]

Good thing we actually validates the extreme value now.

[fsaintjacques]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it? If this is disabled, that should be a blocker for 1.0, parsing (u)int64 seems like a blocking requirement.

[pitrou]

@bkietz one thing, you didn't change anything in Java, so either it was already behaving like this, or the test is disabled for it?

Java probably accepts both forms, because AFAICT the test isn't disabled for it.

[pitrou]

I just looked at the Java code: for Int64 and UInt64, it uses jackson.core.JsonParser::getValueAsString, which accepts both number and string tokens.

[pitrou]

+1

[pitrou]

@sbinet Could you validate the Go changes here?

[fsaintjacques]

I say we move on and accept the go changes.

[wesm]

This patch is failing some of the time in Appveyor due to integer narrowing. I'm fixing the problems in #7352