This is a nice change, and I think making documentation more structured will make it easier to maintain.

I do think it'd be good to replace Table/Row methods with explicit Params / Result methods that make it possible to extract names, types, and structure of parameters. This could be used to generate richer documentation, and support other applications like autocomplete.

That's a good point.

I used "Table" and "Row", since there are few more different table types, and also "Arguments" and "Results" tables are basically the same thing (Arguments have the numbers next to some rows, but only some of them)

The possible table types, by just grepping the source code:

• Arguments
• Examples of output descriptors
• Modes
• Result
• Result with verbosity classifier
  • Result (for verbose = false)
  • Result (for verbose = true)
  • Result (for verbose=false)
  • Result (for verbosity = 0)
  • Result (for verbosity = 1)
  • Result (for verbosity = 2)
  • Result (if verbose is not set or set to false)
  • Result (if verbose is set to true)
  • Result: (for verbose = false)
  • Result: (for verbose = true)
• Result with mode clasifier
  • Result (mode "mallocinfo")
  • Result (mode "stats")

All the Result subtypes are.... yeah basically the same thing, could be treated in some way.

Of course another level would be to somehow match the actual RPC arguments to the arguments in the table, and to the arguments in the short example in the first line, but that is ... just too complex. :)

Good first step would be to replace Table("Arguments") with just Arguments(), the rows with numbered arguments replace with Argument(), add Result("classifier") as a thing, and ... yeah the rest of the tables could stay as Table

The fuctional tests seem to be failing for some reason. I will investigate if I get more general concept ACKs :)

Finally fixed

Concept ACK. The functional tests do seem to be passing for me when I try them locally.

Concept ACK

• Move util files to directory #14555 (Move util files to directory by jimpo)
• Use RPCHelpMan to generate RPC doc strings #14530 (Use RPCHelpMan to generate RPC doc strings by MarcoFalke)
• Add P2SH-P2WSH support to listunspent RPC #14481 (Add P2SH-P2WSH support to listunspent RPC by MeshCollider)
• Add ability to convert solvability info to descriptor #14477 (Add ability to convert solvability info to descriptor by sipa)
• [wallet] Deprecate generate RPC method #14468 ([wallet] Deprecate generate RPC method by jnewbery)
• docs: Consistent type names in RPC help descriptions #14459 (More RPC help description fixes by ch4ot1c)
• Add SegWit support to importmulti #14454 (Add SegWit support to importmulti by MeshCollider)
• Refactor: Start to separate wallet from node #14437 (Refactor: Start to separate wallet from node by ryanofsky)
• [wallet] Restore ability to list incoming transactions by label #14411 ([wallet] Restore ability to list incoming transactions by label by ryanofsky)
• rpcwallet: 'ischange' field for 'getaddressinfo' RPC #14410 (rpcwallet: 'ischange' field for 'getaddressinfo' RPC by mrwhythat)
• Add WalletLocation class #14350 (Add WalletLocation class by promag)
• doc: Fix PSBT howto and example parameters #14319 (doc: Fix PSBT howto and example parameters by priscoan)
• [wallet] Remove addwitnessaddress #14296 ([wallet] Remove addwitnessaddress by jnewbery)
• Index for BIP 157 block filters #14121 (Index for BIP 157 block filters by jimpo)
• Import watch only pubkeys to the keypool if private keys are disabled #14075 (Import watch only pubkeys to the keypool if private keys are disabled by achow101)
• ZMQ: add options to configure outbound message high water mark, aka SNDHWM #14060 (ZMQ: add options to configure outbound message high water mark, aka SNDHWM by mruddy)
• Import key origin data through descriptors in importmulti #14021 (Import key origin data through importmulti by achow101)
• wallet: "avoid_reuse" wallet flag for improved privacy #13756 (wallet: -avoidreuse feature for improved privacy by kallewoof)
• Utils and libraries: Drops the boost/algorithm/string/split.hpp dependency #13751 (Utils and libraries: Drops the boost/algorithm/string/split.hpp dependency by 251Labs)
• refactor: Replace boost::bind with std::bind #13743 (refactor: Replace boost::bind with std::bind by ken2812221)
• wallet/rpc: sendrawtransaction maxfeerate #13541 (wallet/rpc: sendrawtransaction maxfeerate by kallewoof)
• RPC: creates possibility to preserve labels on importprivkey #13381 (RPC: creates possibility to preserve labels on importprivkey by marcoagner)
• wallet: Show fee in results for signrawtransaction* for segwit inputs #12911 (wallet: Show fee in results for signrawtransaction* when known by kallewoof)
• RPC: Add ancestor{count,size,fees} to listunspent output #12677 (RPC: Add ancestor{count,size,fees} to listunspent output by luke-jr)
• RPC: Support addnode onetry without making the connection priviliged #12674 (RPC: Support addnode onetry without making the connection priviliged by luke-jr)
• [rpc] [wallet] Allow specifying the output index when using bumpfee #12096 ([rpc] [wallet] Allow specifying the output index when using bumpfee by kallewoof)
• Optional update rescan option in importmulti RPC #11484 (Optional update rescan option in importmulti RPC by pedrobranco)
• [wallet] [rpc] sendtoaddress/sendmany: Add explicit feerate option #11413 ([wallet] [rpc] sendtoaddress/sendmany: Add explicit feerate option by kallewoof)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

Concept ACK. However, this seems to be a large change that is probably impossible to review without splitting it up in smaller chunks and/or making it a scripted diff.

While the change seems to touch every help text, it doesn't yet make it meaningfully easier to actually produce the help text machine-generated. You split out the rpc method name, but then pass all arguments as a single string and then each of them again as a "Row". This doens't really solve the issue that the documentation is inconsistent in itself (Starting with the white space and formatting of the single line arg string [cf. #14459] and going on to just forgetting the single line arg string [as in importprunedfunds], ...)

Ideally we'd either merge some exhaustive fix similar to #14459 and then a pure scripted-diff to make it auto-generated (and check that the resulting documentation does not differ after the scripted-diff) or we do the auto-generation without the scripted diff and check that the resulting diff in the documentation looks similar to #14459.

I have quickly hacked up a first step of how I think here: #14530

Thanks for feedback.

ad script-diff: I can probably make it script-diffed, it will need some constistency fixed first so the text is parseable, but that can be done as a separate commit. (It is mostly whitespace, missing : after "Arguments" etc)

ad other notes:

Yes, this PR lets most of the arguments/"rows" be as-is, as the various styles seemed too different to me to make consistent, so I gave up and just let it be; I focused mainly on getting rid of the "space-formatting" and making it automatically consistent.

If the goal is indeed to make generation of, for example, the first-line argument etc more automatic, that can be done, but will take more time :) but sure, it will be a nicer code

If I got it correctly. Let's say, walletfundedpsbt

https://bitcoincore.org/en/doc/0.17.0/rpc/wallet/walletcreatefundedpsbt/

The current RPC, at least start, is

walletcreatefundedpsbt [{"txid":"id","vout":n},...] [{"address":amount},{"data":"hex"},...] ( locktime ) ( replaceable ) ( options bip32derivs )

Creates and funds a transaction in the Partially Signed Transaction format. Inputs will be added if supplied inputs are not enough
Implements the Creator and Updater roles.

Arguments:
1. "inputs"                (array, required) A json array of json objects
     [
       {
         "txid":"id",      (string, required) The transaction id
         "vout":n,         (numeric, required) The output number
         "sequence":n      (numeric, optional) The sequence number
       } 
       ,...
     ]
2. "outputs"               (array, required) a json array with outputs (key-value pairs)
   [
    {
      "address": x.xxx,    (obj, optional) A key-value pair. The key (string) is the bitcoin address, the value (float or string) is the amount in BTC
    },
    {
      "data": "hex"        (obj, optional) A key-value pair. The key must be "data", the value is hex encoded data
    }
    ,...                     More key-value pairs of the above form. For compatibility reasons, a dictionary, which holds the key-value pairs directly, is also
                             accepted as second parameter.
   ]

Should the "subcodes" in inputs/outputs arguments, and their help, be also generated from some object (that seems to be m_inner in your example)? Including all the ",...." ? Should the first line stay like it is, or should it instead be walletcreatefundedpsbt inputs outputs ( locktime ) ( replaceable ) ( options bip32derivs )?

I am just saying it will be pretty hard to do this, that's why I opted for "just" adding Row to each table (which also took care of all the Results, that have the same formatting logic). :)

but if the goal is indeed to make all the doc even more uniform, it's probably better than this "naive" approach that just formats the spaces

edit: I look more closely at your #14530 and it indeed produces all the "pseudo-objects" on the left. OK

#14530 looks great, I will try to work on top of that

Closing this, in support of approach similar to #14530

It will be harder to rewrite current source code to that, but ultimately will be more helpful.