Add/improve some tests, especially around size limits by chkno · Pull Request #43 · zowens/commitlog

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno · 2023-03-10T00:47:32Z

Add/improve some tests, especially around message size limits.

After reading the documentation, and the tests, I wasn't clear about what happens when one tries to read a message larger than ReadLimit, so I added a test for it:

so future users can see this at a glance by reading the tests, and
to emphasize that this behavior, down to the exact "Message exceeded max byte size" error message text, is part of the client API: Matching this error message text is the only way for clients to know that they need to retry their read with a larger ReadLimit. I find that I need to match this error text to detect this condition in a thing I'm writing that uses commitlog, and I'd feel better if the specific error text was guarded by a test.
to possibly start a discussion about whether clients ought to be depending on error message text like this / if there should be a better way to signal this 'you need to retry with a larger ReadLimit' condition, (or maybe even a way for a client to know how big the ReadLimit needs to be to read the next message without continually retrying with slightly higher limits). :)

This is a critical part of the read API. Because the default message size write limit (1mb) is larger than the default read size (8kb), clients using the default limits can write messages they cannot read. Clients needs to be able to distinguish this message-too-big-to-read case from other errors so they can retry the read with a larger ReadLimit.

zowens · 2023-03-14T22:14:39Z

@chkno A more specific error message does make sense to me.

chkno added 4 commits March 9, 2023 13:28

Verify round-trip in messagebuf_fromiterator test
7910db6